View MSM66P56_1131902.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

MSM6679A-110 Voice Recognition Processor
MSM6679A-110 Voice Recognition Processor
Semiconductor
Contents
Description .......................................................................................................................................... Features ................................................................................................................................................ Functional and I/O Diagrams .......................................................................................................... Pin Descriptions ................................................................................................................................. Electrical Specifications ..................................................................................................................... Absolute Maximum Ratings ............................................................................................ Operating Conditions ....................................................................................................... DC Characteristics (VDD = 4.5 to 5.5 V, Ta = -40 to 80C) .......................................... AC Characteristics ............................................................................................................ Timing Diagrams .............................................................................................................. Functional Description ...................................................................................................................... Voice Recognition ............................................................................................................. SI Recognition ................................................................................................... SD Recognition ................................................................................................. Name Tag Recording ........................................................................................................ Audio Input Interface ....................................................................................................... Audio Output Interface .................................................................................................... Memory Interface .............................................................................................................. External Voice Synthesis Control ................................................................................... Serial Interface ................................................................................................................... MSM6679A-110 Slave-Mode API .................................................................................................... Command Summary ........................................................................................................ Command Descriptions ................................................................................................... Asynchronous Serial Protocol Example ........................................................................ 1 1 2 8 12 12 12 13 14 15 17 17 17 20 20 21 21 21 24 25 26 27 31 44
2
Semiconductor MSM6679A-110 VRP
Semiconductor
MSM6679A-110 Voice Recognition Processor
SI/SD Voice Recognizer, Recorder/Player, and Speech Synthesizer
DESCRIPTION
The MSM6679A-110 Voice Recognition Processor (VRP) is a slave-mode device that performs five functions: speaker-independent (SI) voice recognition, speaker-dependent (SD) voice recognition, solid-state sound recording, sound playback, and speech synthesis. The highly integrated device also provides an on-chip memory controller, Flash memory interface, analog data conversion, Oki speech synthesizer interface, and pulse width modulation (PWM) sound output. For SI recognition, the MSM6679A-110 contains a vocabulary template in external memory. Pretrained SI vocabularies eliminate the need for laborious training, as usually required by SD products. The memory requirements are dependent on the size of the vocabulary. The MSM6679A110 can tolerate background noise, while providing high recognition accuracy. In its designated operating environment, the device achieves a typical recognition accuracy of >95% (using an Oki-defined test procedure). For SD recognition, the MSM6679A-110 stores SD vocabulary templates, as defined by the user, in external SRAM. The MSM6679A-110 can create SD vocabularies of up to 61 words each, with each word using approximately 50 bytes. In addition to providing voice recognition capabilities, the MSM6679A-110 integrates a solidstate recorder/player, speech synthesis functions, and a tone generator. ADPCM recording/ playback provides high quality sound and efficient memory utilization. The MSM6679A-110 can respond to spoken commands, verbally or with tones, via an on-chip speech synthesizer and tone generator. For larger speech-synthesis requirements, the MSM6679A-110 also provides a glueless MSM665x control interface for off-chip speech synthesis. The MSM6679A-110 can interface to any application or personal computer via a parallel or serial interface through an open, device-independent serial mode API (SMAPI). To accelerate code development, Oki supplies an evaluation kit, and assembly and C language programs for this product.
FEATURES
* SI recognition - Up to 20 - 25 words in each vocabulary - Multiple vocabulary support * SD recognition - Up to 61 words in each vocabulary - Multiple vocabulary support * Speech synthesis - Up to 2.3-sec internal and 27.6-sec external speech synthesis on-chip; sample looping and concatenation allows even longer phrases. - On-chip controller for MSM665x speech synthesizer - Standard beep tone outputs - Pulse code modualation (PCM) and adaptive differential pulse code modualation (ADPCM) voice or soundeffect output * Speech capture and playback - 28-kbps ADPCM speech compression * Serial ASCII command interface * 6944-Hz audio input sample rate for record andplayback * 10-kHz sample rate for voice recognition * 200-msec recognition latency * Flexible memory mapping for EPROM, FLASH, and SRAM * 32-MHz operation * Packages: 84-pin PLCC (QFJ84-P-S115) or 100-pin TQFP (TQFP100-P-1414-0.50-K) 1
MSM6679A-110 Voice Recognition Processor
Semiconductor
FUNCTIONAL AND I/O DIAGRAMS
Analog Input
Recognition and Synthesis Engine
System Controller
Serial Interface PWM
Vocabulary Memory
Algorithm Memory
Output External Speech Synthesis Control
External Memory Control
Figure 1. MSM6679A-110 Block Diagram
ADC0 ~ ADC9 A/D Interface VREF NAR BUSY SI SD STROBE RESOUT VOICEOUT1 RXD1 TXD1 RES OSC0 OSC1
AD0 ~ AD7
A8 ~ A15
Serial-Mode MSM665x Interface
PWM Output
Serial Interface
IC Reset and Oscillator Inputs
ROMRD EA WRRAM RDRAM ALE ES LOADPGM ROMPAGE0 ROMPAGE1 RAMPAGE0 RAMPAGE1 A15FLIP
Memory Interface
Figure 2. MSM6679A-110 Logic Symbol
2
Semiconductor
MSM6679A-110 Voice Recognition Processor
N/C N/C N/C N/C N/C N/C+ RES EA VDD AVDD ADC0 ADC1 ADC2 ADC3 ADC4 ADC5 ADC6 ADC7 ADC8 ADC9 AGND
75 76 77 78 79 80 81 82 83 84 1 2 3 4 5 6 7 8 9 10 11
,
74 73 72 71 70 69 68 67 66 65 64 63 62 61 60 59 58 57 56 55 54
N/C N/C N/C N/C N/C N/C RAMPAGE1 RAMPAGE0 LOADPGM N/C GND SD SI BUSY N/C ROMPAGE1 ROMPAGE0 STROBE A15FLIP N/C N/C
53 52 51 50 49 48 47 46 45 44 43 42 41 40 39 38 37 36 35 34 33
N/C N/C N/C- A15 A14 A13 A12 A11 A10 A9 A8 VDD AD7 AD6 AD5 AD4 AD3 AD2 AD1 AD0 VOICEOUT1
Figure 3. MSM6679A-110 84-Pin PLCC Pinout
VREF N/C+ N/C+ N/C N/C N/C N/C RXD1 TXD1 GND OSC0 OSC1 ALE ROMRD WRRAM RDRAM N/C- RESOUT ES NAR N/C
12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
3
MSM6679A-110 Voice Recognition Processor
Semiconductor
MSM6679A-110 Alphabetic Pin List
Name AD0 AD1 AD2 AD3 AD4 AD5 AD6 AD7 A8 A9 84-pin/ 100-pin 34/28 35/29 36/30 37/31 38/32 39/33 40/34 41/35 43/39 44/40 Name A10 A11 A12 A13 A14 A15 A15FLIP ADC0 ADC1 ADC2 84-pin/ 100-pin 45/41 46/42 47/43 48/44 49/45 50/46 56/54 1/89 2/90 3/91 Name ADC3 ADC4 ADC5 ADC6 ADC7 ADC8 ADC9 AGND ALE AVDD 84-pin/ 100-pin 4/92 5/93 6/94 7/95 8/96 9/97 10/98 11/99 24/16 84/87 Name BUSY N/C- EA ES GND
LOADPGM
84-pin/ 100-pin
Name
84-pin/ 100-pin
Name SI STROBE TXD1 VREF
84-pin/ 100-pin 62/61 57/55 20/10 12/2
61/60 RAMPAGE0 67/67
28,51/20,47 RDRAM
N/C+ 13,14,80/3,4,82 RAMPAGE1 68/68 27/19 81/83 82/84 RES
30/22 RESOUT 29/21 VOICEOUT1 33/27
21,64/12,63 ROMPAGE0
58/56 WRRAM 26/18 VDD
42,83/37,85
66/66 ROMPAGE1 59/57 31/23 22/13 23/15 ROMRD RXD1 SD 25/17 19/9 63/62
NAR OSC0 OSC1
4
Semiconductor
MSM6679A-110 Voice Recognition Processor
Figure 4. MSM6679A-110 84-Pin Package Mechanical Drawing
5
MSM6679A-110 Voice Recognition Processor
Semiconductor
99 AGND
98 ADC9
97 ADC8
96 ADC7
95 ADC6
94 ADC5
93 ADC4
92 ADC3
91 ADC2
90 ADC1
89 ADC0
87 AVDD
82 N/C+
83 RES
81 N/C
80 N/C
79 N/C
78 N/C
77 N/C
100 N/C
76 N/C 75 N/C 74 N/C 73 N/C 72 N/C 71 N/C 70 N/C 69 N/C 68 RAMPAGE1 67 RAMPAGE0 66 LOADPGM 65 N/C 64 N/C 63 GND 62 SD 61 SI 60 BUSY 59 N/C 58 N/C 57 ROMPAGE1 56 ROMPAGE0 55 STROBE 54 A15FLIP 53 N/C 52 N/C 51 N/C N/C 50
88 N/C
86 N/C A9 40
85 VDD A10 41
N/C VREF N/C+ N/C+ N/C N/C N/C N/C RXD1
1 2 3 4 5 6 7 8 9
TXD1 10 N/C 11 GND 12 OSC0 13 N/C 14 OSC1 15 ALE 16 ROMRD 17 WRRAM 18 RDRAM 19 N/C- 20 RESOUT 21 ES 22 NAR 23 N/C 24 N/C 25 AD0 28 AD1 29 AD2 30 AD3 31 AD4 32 AD5 33 AD6 34 AD7 35 N/C 36 VDD 37 N/C 38 A8 39 A11 42 A12 43 A13 44 A14 45 A15 46 N/C- 47 N/C 48 N/C 49
N/C 26
VOICEOUT1 27
Figure 5. MSM6679A-110 100-Pin TQFP Pinout
6
84 EA
Semiconductor
MSM6679A-110 Voice Recognition Processor
Figure 6. MSM6679A-110 100-Pin Package Mechanical Drawing
7
MSM6679A-110 Voice Recognition Processor
Semiconductor
PIN DESCRIPTIONS
Pin # 84-pin/ Pin Name 100-pin -/1 1/89 2/90 3/91 4/92 5/93 6/94 7/95 8/96 9/97 10/98 N/C ADC0 ADC1 ADC2 ADC3 ADC4 ADC5 ADC6 ADC7 ADC8 ADC9 Analog Ground. This pin provides an analog ground point, allowing 11/99 AGND Analog ground independent grounding of the analog and digital circuitry. Separate grounds reduce the impact of digital switching noise on analog sampling accuracy. Analog Reference Voltage. The MSM6679A-110's on-chip A/D 12/2 13/3 14/4 15/5 16/6 17/7 18/8 19/9 20/10 -/11 21/12 VREF N/C+ N/C+ N/C N/C N/C N/C RXD1 TXD1 N/C GND Input Output (not connected) Ground Serial Port Receive. This is the receive data line for serial port. Serial Port Transmit. This is the transmit data line for serial port. Reserved. This pin is reserved for future use and should be left open. Ground. Oscillator 0/External Clock. When the MSM6679A-110 uses a crystal 22/13 OSC0 Input oscillator, this input is the oscillator input pin. The pin is then connected to one side of a crystal and load capacitor. When used with an external clock, the external clock is applied to this input. -/14 N/C (not connected) Reserved. This pin is reserved for future use and should be left open. Oscillator 1. When the MSM6679A-110 uses a crystal oscillator, this 23/15 OSC1 Output output is the oscillator output pin. The pin is then connected to one side of a crystal and load capacitor. When used with an external clock, this output is left unconnected. 24/16 ALE Output Memory Address Latch Enable. An external memory latch is controlled by this signal, the address latch enable output. (do not connected) Reserved. These pins are reserved for future use and must be left open. Reference voltage converter uses this analog reference voltage when converting an analog signal into digital samples Input Reserved. These pins are reserved for future use and must be tied to VDD. Analog input Analog Input. These ten inputs are tied together and serve as the analog input. Signal conditioning, via a bandpass filter and gain circuit, is required before this input. (not connected) Reserved. This pin is reserved for future use and should be left open. Signal Type Description
8
Semiconductor
MSM6679A-110 Voice Recognition Processor
Pin # 84-pin/ Pin Name 100-pin ROM Read. This is a strobe signal for direct connection to an external 25/17 ROMRD Output ROM's READ input. When asserted LOW, this signal indicates that the MSM6679A-110 is ready to read data from the ROM. RAM Write. This is a strobe signal for direct connection to an external 26/18 WRRAM Output RAM's WR input. When asserted LOW, this signal indicates that the MSM6679A-110 is ready to write data to RAM. RAM Read. This is a strobe signal for direct connection to an external 27/19 28/20 29/21 30/22 RDRAM N/C- RESOUT ES Output Input Output Output RAM's RD input. When asserted LOW, this signal indicates that the MSM6679A-110 is ready to read data from RAM. Reserved. This pin is reserved for future use and must be tied to GND. MSM665x Reset. This pin provides a reset signal for an external speech synthesis engine. Flash Bank Control (Extended Segments). This is the control signal for flash memory banking. MSM665x Next Address Request. This pin signals to the 31/23 32/24 -/25,26 33/27 34/28 35/29 36/30 37/31 38/32 39/33 40/34 41/35 -/36 42/37 -/38 43/39 44/40 45/41 46/42 47/43 48/44 NAR N/C N/C VOICEOUT1 AD0 AD1 AD2 AD3 AD4 AD5 AD6 AD7 N/C VDD N/C A8 A9 A10 A11 A12 A13 Outputs Memory Address Bus. These are the upper eight address pins. (not connected) Digital Power (not connected) Reserved. This pin is reserved for future use and should be left open. Power. Reserved. This pin is reserved for future use and should be left open. Bidirectional I/O Memory Address/Data Bus. These are multiplexed address/data lines for the eight data bits and the lower eight address bits (the upper eight address bits are not multiplexed). Input (do not connect) (not connected) Output MSM6679A-110 that the external speech synthesis engine is ready for another command. Reserved. These pins are reserved for future use and must be left open. Reserved. These pins are reserved for future use and should be left open. Voice Out. This pin is the PWM output for speech synthesis, voice sample playback, and voice prompts. An external integrator must be used to convert this to an analog signal. Signal Type Description
9
MSM6679A-110 Voice Recognition Processor
Semiconductor
Pin # 84-pin/ Pin Name 100-pin 49/45 50/46 51/47 52/48 53/49 -/50,51 54/52 55/53 56/54 57/55 58/56 59/57 60/58 -/59 61/60 A14 A15 N/C- N/C N/C N/C N/C N/C A15FLIP STROBE ROMPAGE0 ROMPAGE1 N/C N/C BUSY Outputs Input (do not connect) (not connected) (do not connect) Output Output Outputs (do not connect) (not connected) Input Memory Address Bus. These are the upper eight address pins. Reserved. This pin is reserved for future use and must be tied to GND. Reserved. These pins are reserved for future use and must be left open. Reserved. These pins are reserved for future use and should be left open. Reserved. These pins are reserved for future use and must be left open. Memory Address A15 Flip. This signal inverts the A15 address signal for 32-Kbyte bank switching on the local memory bus. MSM665x Strobe. This output provides the LOAD signal for an external speech synthesizer. ROM Page Select. These signals select one of four 64-Kbyte ROM pages. Reserved. This pin is reserved for future use and must be left open. Reserved. This pin is reserved for future use and should be left open. MSM665x Busy. When using an external MSM665x device, this pin monitors the MSM665x BUSY signal and connects directly to the MSM665x BUSY signal output. 62/61 63/62 64/63 -/64 65/65 66/66 67/67 68/68 69/69 70/70 71/71 72/72 73/73 74/74 -/75,76 SI SD GND N/C N/C LOADPGM RAMPAGE0 RAMPAGE1 N/C N/C N/C N/C N/C N/C N/C (not connected) Reserved. These pins are reserved for future use and should be left open. (do not connect) Reserved. These pins are reserved for future use and must be left open. Output Output Digital Ground (not connected) (do not connect) Output MSM665x Serial Clock. This MSM6679A-110 output connects to the MSM665x SI input. The SI pin is the MSM665x serial clock input pin. MSM665x Serial Data. This MSM6679A-110 output connects to the MSM665x SD input. The SD pin is the MSM665x serial data input pin. Ground. Reserved. This pin is reserved for future use and should be left open. Reserved. This pin is reserved for future use and must be left open. Load Program. This signal allows the MSM6679A-110 to write data to program memory. When asserted low, this signal should set the program memory in write mode. Output RAM Page Select. These signals support selection of one out of four RAM pages. Each page is 64kbytes in size. Signal Type Description
10
Semiconductor
MSM6679A-110 Voice Recognition Processor
Pin # 84-pin/ Pin Name 100-pin 75/77 76/78 77/79 78/80 79/81 80/82 81/83 N/C N/C N/C N/C N/C N/C+ RES Input Input Reserved. This pin is reserved for future use and must be tied to VDD. MSM6679A-110 Reset. External logic should assert this power-on reset signal LOW when power is applied to the MSM6679A-110. External ROM Address Select. This control signal enables external 82/84 83/85 -/86 84/87 -/88 -/100 EA VDD N/C AVDD N/C N/C Input ROM execution. This signal is usually connected to ROMPAGE1 and a pullup resistor. Positive digital supply Power. (not connected) Analog power supply (not connected) (not connected) Reserved. This pin is reserved for future use and should be left open. Analog Power. Reserved. This pin is reserved for future use and should be left open. Reserved. This pin is reserved for future use and should be left open. (do not connect) Reserved. These pins are reserved for future use and must be left open. Signal Type Description
11
MSM6679A-110 Voice Recognition Processor
Semiconductor
ELECTRICAL SPECIFICATIONS
Absolute Maximum Ratings
Parameter Digital power supply voltage Input voltage Output voltage Analog power voltage Analog reference voltage Analog input voltage Power dissipation Storage temperature Symbol VDD VI VO AVDD VREF VAI PD TSTG Ta = 85C, per package Ta = 85C, per pin -- GND = AGND = 0 V Conditions Value -0.3 to +7.0 -0.3 to VDD +0.3 -0.3 to VDD +0.3 -0.3 to VDD +0.3 -0.3 to AVDD +0.3 -0.3 to VREF 1300 max. 50 max. -50 to +150C mW C V Unit
1.
Permanent device damage may occur if ABSOLUTE MAXIMUM RATINGS are exceeded. Functional operation should be restricted to the conditions as detailed elsewhere in this data sheet. Exposure to absolute maximum rating conditions for extended periods may affect device reliability.
Operating Conditions
Parameter Digital power supply voltage Analog power supply voltage Analog reference voltage Analog input voltage Storage holding voltage Operating frequency Ambient temperature Fan-out Symbol VDD AVDD VREF VAI VDDH fOSC Ta N fOSC = 0 MHz VDD = 5 V 10% -- MOS load TTL load, AD0 ~ AD7 TTL Load, all other outputs Conditions fOSC = 32 MHz VDD = AVDD Value 4.5 to 5.5 4.5 to 5.5 AVDD -0.3 to AVDD AGND to VREF 2.0 to 5.5 32 -40 to 85C 20 2 1 MHz C V Unit
12
Semiconductor
MSM6679A-110 Voice Recognition Processor
DC Characteristics (VDD = 4.5 to 5.5 V, Ta = -40 to 85C)
Parameter Symbol Condition Applied to AD0-AD7 High-level input voltage VIH Applied to OSC0 Applied to all other I/O Applied to AD0-AD7 Low-level input voltage VIL Applied to OSC0 Applied to all other I/O Output current = 400 mA, applied High-level output voltage VOH to AD0-AD7, ALE, and ROMRD Output current = 200 mA, for all other I/O Output current = 3.2 mA, applied Low-level output voltage VOL to AD0-AD7, ALE, and ROMRD Output current = 1.6 mA, for all other I/O Input leak current IIH, IIL Input current High-level output current Low-level output current Output leakage current Input capacitance Output capacitance Analog reference power supply voltage Power consumption IOH IOL ILO CI CO IREF IDD VI = VDD/0 V, applied to Ain, EA, FLOAT, and RESTART VI = VDD/0 V, applied to RES VI = VDD/0 V, applied to OSC0 VO = 2.4 V, applied to AD0-AD7 VO = 2.4 V, applied to all other I/O VO = 2.4 V, applied to AD0-AD7 VO = 2.4 V, applied to all other I/O VO = VDD/0 V f = 1 MHz, Ta = 25C During voice input When voice input is halted fOSC = 32 MHz, no load -- -- -- -- -- Rated Value Min 2.2 0.85 x VDD 0.80 x VDD -0.3 -0.3 -0.3 VDD -0.4 VDD -0.4 -- -- -- -- -- -2 -1 10 5 Typ [1] -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 5 7 -- -- 55 Max VDD +0.3 VDD +0.3 VDD +0.3 0.8 0.15 x VDD 0.2 x VDD -- -- 0.4 0.4 1/-1 1/-250 15/-15 -- -- -- -- 2 -- -- 4 10 75 A pF mA A mA mA A V Unit
1.
Typical condition is 5 V 25C.
13
MSM6679A-110 Voice Recognition Processor
Semiconductor
AC Characteristics External Program Memory Control (VDD = 4.5 to 5.5 V, Ta = -40 to 85C)
Parameter Clock pulse width (OSC) ALE pulse width ROMRD pulse width ROMRD pulse delay time Low address set-up time Low address hold time High address delay time High address hold time Instruction set-up time Instruction hold time Symbol tOW tAW tPW tPAD tAAS tAAH tAAD tAPH tIS tIH CL = 50 pF Condition -- Min. 15.625 36.875 177.5 10.625 21.25 10.625 15.625 15.625 35 0 Max. -- -- -- 20.625 41.25 20.625 25.625 25.625 -- 25.625 ns Unit
External Data Memory Control (VDD = 4.5 ~ 5.5 V, Ta = -40 ~ 85C)
Parameter Clock pulse width (OSC) ALE pulse width RDRAM pulse width WRRAM pulse width RDRAM pulse delay time WRRAM pulse delay time Low address set-up time Low address hold time High address set-up time High address hold time Memory data set-up time Memory data hold time Data set-up time Data hold time Symbol tOW tAW tRW tWW tRAD tWAD tAAS tAAH tAAD tARH, tAWH tMS tMH tDD tDH CL = 50 pF Condition -- Min. 15.625 36.875 177.5 177.5 10.625 10.625 21.25 10.625 15.625 15.625 35 0 15.625 15.625 Max. -- -- -- -- 20.625 20.625 41.25 20.625 25.625 25.625 -- 5.625 25.625 25.625 ns Unit
14
Semiconductor Timing Diagrams
tOW CLK tAW ALE tOW
MSM6679A-110 Voice Recognition Processor
tPAD ROMRD tAAS AD0 - AD7 tAAD A8 - A15 PC8 - 15 PC0 - 7 tAAH
tPW
tIS
tIH
INST0 - 7 tAPH
Figure 7. ROM Read Timing
tOW CLK
tOW
tAW ALE tRAD RDRAM tAAS AD0 - AD7 tAAD A8 - A15 RAP8 - 15 RAP0 - 7 tAAH tMS tMH tRW
INST0 - 7 tARH
Figure 8. RAM Read Timing 15
MSM6679A-110 Voice Recognition Processor
Semiconductor
tOW CLK
tOW
tAW ALE tWAD WRRAM tDD tAAS AD0 - AD7 tAAD A8 - A15 RAP8 - 15 RAP0 - 7 tAAH DOUT0 - 7 tAWH tDH tWW
Figure 9. RAM Write Timing
16
Semiconductor
MSM6679A-110 Voice Recognition Processor
FUNCTIONAL DESCRIPTION
Voice Recognition The MSM6679A-110 performs both SD and SD recognition. SI vocabularies are embedded in the MSM6679A-110. For SD recognition, each recognized phrase must be enrolled in the MSM6679A110's vocabulary by creating a composite template from multiple recordings of the same phrase. Then the composite tempalte is stored in SRAM or FLASH memory. During both SI and SD recognition, the MSM6679A-110 performs the following steps: 1. After external band-pass filtering, the MSM6679A-110 converts the analog signal to PCM samples. 2. The MSM6679A-110 extracts significant features from the sample data by frequency and time-domain analysis. 3. The MSM6679A-110 compares the analyzed input with the reference data for each signal, weighing the significance of similarities according to control software parameters. A score (expressed as distance) is generated for each phrase. 4. he vocabulary phrase that achieves the highest score (or lowest distance) is judged to match the input phrase, assuming that the score exceeds a predetermined threshold. 5. Via a special command, the MSM6679A-110 can also return the scores of the input against all defined vocabulary phrases for SI or SD recognition. This feature allows external host software to select the next best match, if the closest match is not contextually logical.
SI Recognition
Oki supplies the MSM6679A-110 with predefined SI vocabularies which Oki builds from hundreds of utterances by a wide variety of speakers. SI vocabularies are limited to 25 words or less, which allows the MSM6679A-110 to achieve a net accuracy of >95%, even in noisy conditions. SI vocabularies are grouped into sub-vocabularies of 15 words, to maintain the highest accuracy. Similar words in any one sub-vocabulary can cause substitution errors. Oki Semiconductor's standard cellular vocabulary is intended for an automotive environment with a far-talk microphone. This vocabulary may work adequately in other conditions, such as an office or outside, but recognition performance may be degraded. MSM6679A-110 Cellular SI Recognition Vocabulary
Sub-Vocabulary 1 Phrase Store Dial Delete Directory -- -- -- Index 1 2 3 4 -- -- -- Phrase One Two Three Four Five Six Seven Sub-Vocabulary 2 Index 1 2 3 4 5 6 7 Phrase Eight Nine Zero Oh Stop Clear -- Index 8 9 Ah Bh Ch Dh -- Sub-Vocabulary 3 Phrase Yes No Cancel -- -- -- -- Index 1 2 3 -- -- -- --
17
MSM6679A-110 Voice Recognition Processor
Semiconductor
MSM6679A-110 Control Vocabulary
Sub-Vocabulary 1 Sub-Vocabulary 2 Phrase A/C Fan Temperature Timer Service Help Select -- -- -- -- -- Index 1 2 3 4 5 6 7 -- -- -- -- -- Phrase Low Medium High Increase Decresse Set Reset Cancel Clear Recall On Help Index 1 2 3 4 5 6 7 8 9 A B C
MSM6679A-110 Direction Vocabulary
Sub-Vocabulary 1 Phrase Up Down Left Right Formard Reverse Faster Slower Start Stop Cancel Index 1 2 3 4 5 6 7 8 9 A B
MSM6679A-110 Browse Vocabulary
Sub-Vocabulary 1 Phrase Up Down Left Right Index 1 2 3 4 Phrase Next Previous Select Cancel Index 5 6 7 8 Phrase Home -- -- -- Index 9 -- -- -- Phrase Set Reset Start Stop Sub-Vocabulary 2 Index 1 2 3 4 Phrase On Play Lock Cancel Index 5 6 7 8
18
Semiconductor
MSM6679A-110 Voice Recognition Processor
MSM6679A-110 Japanese Navigation Vocabulary
Sub-Vocabulary 1 Sub-Vocabulary 2 Sub-Vocabulary 3 Sub-Vocabulary 4 Phrase Genzaichi Jiaku Kaisya Houi Sentaku Yuudou Nabi -- Index 1 2 3 4 5 6 7 -- Phrase Ue Shita Hidari Migi -- -- -- -- Index 1 2 3 4 -- -- -- -- Phrase Hyoujun Kakudai Shukushou Zentai Kaiten Kyori Hosei Teisei Index 1 2 3 4 5 6 7 8 Phrase Hai Iie Ofu -- -- -- -- -- Index 1 2 3 -- -- -- -- --
MSM6679A-110 Japanese Celluar Vocabulary
Sub-Vocabulary 1 Phrase On Ofu Daiyaru Tansyuku Denwacho Kakunin Nabi -- Index 1 2 3 4 5 6 7 -- Phrase Ichi Ni San Yon Go Roku Nana Hachi Sub-Vocabulary 2 Index 1 2 3 4 5 6 7 8 Phrase Kyuu Zero Sharp Star Kakunin Touroku Rei -- Index 9 A B C D E F --
SI vocabulary generation starts with collecting reference utterances from 400 speakers with: * An equal mixture of males and females * Accents from all regions of the country of intended use * ~15% non-native speakers. The samples should be generated from a randomly-ordered list, with each word spoken twice and with a dummy word at the beginning and end. There must be >2 sec between each sample for accurate data processing. To provide the audio fidelity required for high-quality recognition training, a DAT recorder, together with the microphone that will be used in the final application, is required. To ensure data integrity, data is submitted to Oki after collecting samples from the first 20 speakers for initial screening. If acceptable, then the remaining collection may proceed. If substitution errors are possible, collection of spare words during initial collection is recommended. For example, alternate words to "Stop" and "Top" could be "Halt" and "First." Collections should contain a wide variety of the background sound conditions that will exist during actual usage. For example, if the collection is for use in an automobile, conditions such as vehicle speed, road conditions, various window opening positions, heater or AC blower speeds and radio volumes should be varied during the collection. The signal-to-noise ratio should be maintained at 20dB.
19
MSM6679A-110 Voice Recognition Processor
Semiconductor
To achieve high accuracy rates, phrase selection, data collection, background initialization strategy, and control software need careful consideration. There are no published standards for recognition accuracy. Oki defines accuracy by: Accuracy = 100% - ERATE ERATE = ESUB + 1/2 EREJ with the following definitions: Parameters for Recognition Accuracy
Name Substitution Error Rejection Error Gap Error Time-Out Error Spurious Response Error Symbol ESUB EREJ EGAP ETME ESPU Condition Most critical type error, e.g., Say "Five", recogrize "Nine" Word not recognized, opportunity for operator to repeat Word spoken before recognizer ready Word length is too long Sourd or imvalid word classfied as a valid word (i.e., drop handset or speak wong word)
A typical target accuracy of 97% is achieved with a 3% ERATE , composed of a 1.5% ESUB rate and a 3%EREJ rate.
SD Recognition
In SD recognition mode, the MSM6679A-110 can be trained to recognize up to 61 words. The MSM6679A-110 can support multiple speakers by switching vocabularies, but only one speaker's vocabulary should be active at one time. The end user enrolls a phrase in the MSM6679A-110's vocabulary by recording the phrase three times or more. The host Micro Controller Unit (MCU) controls the number of times each phrase in enrolled. Generally, higher recognition accuracy is achieved with each additional enrollment. The word set is made more robust by pronouncing each phrase slightly differently during initial enrollment. In addition to enrollment training, adaptive template updating can drive the accuracy towards 100%. The host MCU updates templates by first asking the speaker to confirm a recognized phrase with a "yes" or "no" response, and subsequently updating the template for corresponding words. The use of name tags (see next paragraph) facilitates this process. Name Tag Recording To facilitate SD recognition, the MSM6679A-110 supports recording and playback of name tags. Name tags are used to confirm correct responses in SD recognition. For example, in a phone dialer application, the user associates a "name" (which is recorded into memory) with a phone number. The MSM6679A-110 then plays back the name tag so that the user can verify that the recognized phrase is the correct one. The VRP stores names tags in memory using an ADPCM compression algorithm with 28 kbps of speech. The length of a name tag is controlled with a command from the users host MCU program. The maximum number of name tags possible is 61, but the actual number is dependent upon record time and memory available. See the section on memory interface for more detail.
20
Semiconductor
MSM6679A-110 Voice Recognition Processor
Audio Input Interface
A critical item for high-accuracy speech recognition is correct design of the audio input circuit. A circuit with appropriate gain and frequency responses must be placed between the microphone and MSM6679A-110's A/D input. Oki recommends input gain and a band pass filter with the following characteristics: * Four pole Chebyshev high-pass filter, 3 dB point at 225 Hz * Dual-pole low-pass filter, 3 dB point at 4250 Hz * Midband gain of 46 dB at 1000 Hz The above gain and filter characteristics are obtained by using a rail-to-rail quad CMOS op-amp and one-half supply rail splitter to bias the input signal at 2.5 V nominal. The MSM6679A-110 uses multiple analog inputs to improve sampling quality. An on-chip analogy to digital (A/D) conversion unit transforms the analog signal to a digital data stream.
Audio Output Interface
The MSM6679A-110 also provides the VOICEOUT1 PWM output. The MSM6679A-110 uses ADPCM to generate voice or sound-effect output. ADPCM represents an improvement over conventional PCM techniques in that it adaptively changes the quantizer step (scale factor) to suit the waveform being encoded. The result is more efficient memory usage with no loss of quality. Careful selection of the components for internal and external output filters and amplifiers is recommended. An incorrect choice would impair the original quality. This consideration equally includes: * Careful separation of analog and digital lines * Grounding of analog lines at both ends * Further adequate separation from high-speed digital circuits to avoid distortions thereof Memory Interface The memory control section manages RAM and/or ROM devices in two 64-Kbyte memory spaces, in conjunction with internal memory for voice templates and working memory. Some versions work with no external memory, some have some external RAM, some use only external EPROM, and some use external memory in conjunction with both internal ROM and RAM. The MSM6679A-110 requires a minimum of 32 Kbytes SRAM and 16 Kbytes ROM. The following table shows vocabulary sizes and playback facilities for various configurations.
21
MSM6679A-110 Voice Recognition Processor
Semiconductor
Typical Configurations
Recognition MSM6679A-110 Application Vocabulary Sound Playback (Words) SI Controller 25 50 25 Telephone Dialer Computer Peripheral Minimum Configuration 50 75 100 61[3] 12 SD 61[2] 61[2] 61 61 61 61 61 61[2]
(sec) [1] Internal External MSM665x MSM6679A-110 MSM6679A-110 Playback Interface
Memory Size (bytes)
EPROM Flash SRAM
Speech Record -- -- OK OK OK OK -- --
Speech Playback OK OK OK OK OK OK OK -- -- 16K --
2.3 2.3 2.3 2.3 2.3 2.3 2.3 1.15
9.2 -- 27.6 18.4 -- -- 36.8
OK OK OK OK OK OK OK OK
64K
--
32K
128K
32K
-- --
64-384K 32K
1. 2. 3.
Phrase chaining features usually permit much longer overall playback durations; not including external speech synthesizer. SD recognition vocabularies are volatile in these configurations. Per download. Vocabulary swapping by host permits unlimited vocabulary size.
The MSM6679A-110 supports up to 64 Kbytes of RAM per bank, and up to 64 Kbytes of ROM per bank in separate memory spaces. The 8-bit data bus is multiplexed with the lower eight address bits; the upper eight address bits are not multiplexed. To demultiplex the address and data bits during all read and write cycles, the MSM6679A-110 requires an external octal latch, such as the 74H373. The MSM6679A-110's Address Latch Enable (ALE) signal controls the octal latch. For accessing the ROM and RAM address spaces, the MSM6679A-110 provides the separate Write RAM (WRRAM), Read RAM (RDRAM), and ROM Read (ROMRD) signals. The RDRAM and ROMRD signals connect directly to Output Enable (OE) control signal inputs on the RAM and ROM, respectively. The WRRAM signal connects directly to the Write Enable (WE) control signal input on the RAM. The following diagrams show the memory maps for the MSM6679A-110. In all MSM6679A-110 memory maps, the DL data memory space must be in RAM. The DH data memory space and PH program memory space can either be implemented in ROM, EPROM, FLASH, RAM, or PROM. In standalone applications, flash memory can be used for recording and subsequent playback of voice prompts (e.g., the user's name) and user sounds (e.g., DTMF dial tones, etc.). Figure 10 shows the configuration for writing to flash memory used when writing SD templates or when flash is used for data memory.
22
Semiconductor
MSM6679A-110 Voice Recognition Processor
Program Memory
00000
FLASH
00000 03FFF 04000 07FFF 08000 08FFF 0C000 0FFFF 10000 13FFF 14000 17FFF 18000 18FFF 1C000 1FFFF
Data Memory
00000 00000
SRAM SO
07FFF 08000
MSM6679 Internal Program Memory
BFFF C000 FFFF
F0 F1 F2 F3 F4 F5 F6 F7
DL
07FFF 08000
DH
0FFFF 0FFFF 10000
S1
PHH
S2
17FFF 18000
Hidden
S3
1FFFF
Figure 10. MSM6679A-110 Program/Data Memory Map (LOADPGM = "0") Figure 11 shows the memory map during all other modes of operation.
Program Memory
00000
FLASH
00000 03FFF 04000 07FFF 08000 08FFF 0C000 0FFFF 10000 13FFF 14000 17FFF 18000 18FFF 1C000 1FFFF
Data Memory
00000
SRAM
00000 07FFF 08000 0FFFF 10000 17FFF 18000 1FFFF
MSM6679 Internal Program Memory
BFFF C000 FFFF
F0 F1 F2 F3 F4 F5 F6 F7
S0 S1 S2 S3
DL
07FFF 08000
DH
0FFFF
PHH
Figure 11. MSM6679A-110 Program/Data Memory Map (LOADPGM = "1") Figure 12 shows the details of the external memory allocation of the MSM6679A-110.
23
MSM6679A-110 Voice Recognition Processor
Semiconductor
FLASH 00000 SI First (F509*) 07300 SD First 07D80 NTP First 08000 10000 Name Tag Data 18000 SI Last (F501*) 1F900 SD Last 1FD80 NTP Last 1FFFF *Denotes commands to select blocks 200 18000 Name Tag Block Address 000 100
FLASH Reserved 04AD0 Default Working SD Templates 05480 Working Name Tag Pointer Table 05700 Alternate SD Templates 08000 Down load RAM Bank 0F300 (F510*) Alternate SD Templates 10000 Reserved
Buffer RAM Bank (F520*) 2F6 2FB 2FF 1F300 1FFFF Reserved
Figure 12. MSM6679A-110 External Memory Map External Voice Synthesis Control The MSM6679A-110 is capable of interfacing to the MSM665x family of Oki ROM, OTP, or external EPROM speech synthesizers, allowing for up to 260 seconds of high-quality voice and sound effects. The following table indicates the speech capabilities of the MSM665x family. MSM665x Family Characteristics
Type MSM6650 MSM6652 MSM6653 MSM66P54[4] MSM6654 MSM6655 MSM66P56[5] MSM6656 MSM6658 Data ROM 64 Mbits[3] 288 Kbit 544 Kbit 1 Mbit 1 Mbit 1.5 Mbit 2 Mbit 2 Mbit 4 Mbit >1 hour 16.9 sec 31.2 sec 63.8 sec 63.8 sec 96.5 sec 129.1 sec 129.1 sec 258 sec Maximum Speech Duration[2] >40 minutes 10.5 sec 19.5 sec 39.9 sec 39.9 sec 60.3 sec 80.7 sec 80.7 sec 161.4 sec >30 minutes 8.4 sec 15.6 sec 31.9 sec 31.9 sec 48.2 sec 64.5 sec 64.5 sec 129.1 sec >15 minutes 4.2 sec 7.8 sec 15.9 sec 15.9 sec 24.1 sec 32.2 sec 32.2 sec 64.5 sec >8 minutes 2.1 sec 3.9 sec 7.9 sec 7.9 sec 12.0 sec 16.1 sec 16.1 sec 32.2 sec
Capacity[1] fSAM = 4.0 kHz fSAM = 6.4 kHz fSAM = 8.0 kHz fSAM = 16.0 kHz fSAM = 32.0 kHz
1.
Actual ROM area in MSM6652, MSM6653, MSM6654, MSM6655, and MSM6656, MSM6658, MSM66P54, MSM66P56 is smaller by 22 Kbits.
24
Semiconductor 2. 3. 4. 5.
MSM6679A-110 Voice Recognition Processor
Longer speech patterns can be created by chaining and repeating existing speech samples. Via external ROM only (no on-chip ROM available). One-Time-Programmable (OTP) version of MSM6654. See the MSM66P54 data sheet for more information. One-Time-Programmable (OTP) version of MSM6656. See the MSM66P56 data sheet for more information.
The MSM665x interface consists of the following signals: * BUSY - Asserted LOW during MSM665x device playback. The MSM6679A-110 F50Bh and F10100xxh commands select this signal for MSM665x command polling. * NAR - Next Address Request status signal. By default, the MSM6679A-110 uses this signal to poll commands to the MSM665x. The F51Bh, F480h, and F440h commands select NAR for polling. * SI - Serial Input Clock. * SD - Serial Data Out. * STROBE - Initiates speech synthesis. * RESOUT - Initializes device when asserted LOW. The MSM6679A-110 F480h command generates this signal. Serial Interface The MSM6679A-110 supplies a serial interface suitable for connection to an RS-232C serial port buffer or equivalent. The serial interface uses one MSM6679A-110 input (RXD) and one MSM6679A-110 output (TXD). The interface operates at 9600 Baud with: * 8 data bits * 1 start bit * 1 stop bit * No parity * No handshake A host processor sends serial ASCII commands to the MSM6679A-110 and receives serial ASCII responses based on voice input responses.
25
MSM6679A-110 Voice Recognition Processor
Semiconductor
MSM6679A-110 SLAVE-MODE API
This section describes the slave-mode Applications Protocol Interface (API) between a host MCU and the MSM6679A-110. The slave-mode API offers the following features: * Direct slave-mode control voice recognition, sound recording and playback, and sound synthesis * Serial port interfaces * Simple procedures for downloading and uploading data * ASCII format * Comprehensive return codes and error reporting The host MCU selects the active speech recognition vocabulary, speech responses, and controls all actions required to implement an interactive voice response system. The MSM6679A-110 performs speech recognition, based on the vocabulary selected by the host, and returns digital codes representing the most probable match of the current utterance to an individual utterance in the selected vocabulary. The MSM6679A-110 can also respond with "name tags." Name tags can be fixed words, phrases or sound effects, or can be words, phrases or sound effects that have been interactively recorded by the user. The API supports both serial and parallel interfaces. The MSM6679A-110 returns each response using the same interface through which the most recent message was received. The user can thus connect and use both interfaces. For all messages, the parallel interface uses 8-bit binary values, while the serial interface represents each 8-bit value with two hexadecimal digits coded in ASCII. When downloading and uploading data, the MSM6679A-110 uses a stream of 8-bit binary values on both parallel and serial interfaces. The serial-mode interface uses a 9600-baud UART with 1 start bit, 8 data bits, and 1 stop bit. There is no parity or handshaking. Serial-interface messages are of variable length, but consist of an even number of bytes. The serial interface echoes all received ASCII characters immediately back to the host MCU. Messages are of variable length. All messages consist of an even number of bytes. Opcodes consist of exactly four bytes, with values between F000h and FEFEh. Operand bytes may take values from 0000h to FFFFh. The MSM6679A-110 issues a return code for many of the host commands. The return code generally consists of the same opcode, followed by data indicating success of failure of the operation. Opcodes are organized into the following categories: * Purge * Set parameter * Initialize * Recognize * Speak * Request * Record * SD recognition control The following tables summarize available opcodes and provide detailed descriptions of the opcode functions.
26
Semiconductor
MSM6679A-110 Voice Recognition Processor
Command Summary
Function Purge Opcode (Hex) F000 F102 xxxx F103 xxxx F104 xxxx F11x F12x F130 xxxx F440 F2xx mod 80 F2xx mod 40 F2xx mod 20 F2xx mod 10 F2xx mod 8 F2xx mod 4 F2xx mod 2 F2xx mod 1 F300 F301 to F33F F340 F341 F342 F343 F344 F351 F361 F371 F401 to F43D F441 to F47C F47E F47F F480 F481 - F4FF F50B F51B FE03 to FEFE F500 F501 F510 Request F520 F522 F513 Description Clear MSM6679A-110 input stack Set SP/SI origin to xxxx. Set SD origin. Set triggering origin. Set IRQ level to IRQ x. Set SD SP table to table x. Select triggering table. Set ISA mode. Initialize background estimation. Wait for F3h command after each response. Beep after each triggered utterance Reserved Set speech response level to default. Send acknowledge after each speech output response. Only detect triggers. Initialize SD parameter table and name tags. Stop listening (recognition). Start SI recognition. Start SD recognition. Sort SD recognition distances, return index to utterance with least distance. Update SD enrollment. Request recognition parameter upload to host. Sort SD recognition distances, return index and distance to utterance with least distance Sort SD recognition distances, return all distances. Sort SD recognition distances, return minimum and maximum energy values. Sort SD recognition distances, return all energy values and distances. Play back name tag from external memory. Play back sound from internal memory. Play 50-ms beep. Pause for 0.2 sec. Initialize MSM665x IC, set MSM665x busy mode OFF, select FLASH SI recognition. Play back one of 127 phrases in external MSM665x device. Set MSM665x busy mode ON. Set 6654 NAR mode Set output volume (03h = minimum, FEh = maximum). Status request. Select last FLASH bank for SI recognition. Select download RAM bank for speaker independent/signal processing (SI/SP) template area. Select buffer RAM bank for SI/SP. Copy download RAM bank to buffer RAM bank Save download RAM bank templates in first FLASH. (8000 - F2FF) Default (Hex) -- 8000 4A00 F100 0005 F123 0101, 0202... Disabled. Disabled. Disabled. Disabled. Disabled. Enabled. Enabled. Disabled. Load from first FLASH. -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- OFF ON FE80h -- F509 F509 -- -- --
Set parameter
Initialize
Recognize
Speak
27
MSM6679A-110 Voice Recognition Processor
Semiconductor
Function
Opcode (Hex) F514 F515 F516 F502.... F503 xxxx F504 F505 F506 F507 F517 F508 F518 F509 F101 00xx F105 F106 F50A F50C F51C F50D F51D F50E F50F FA01 ~ FA3D
Description Get download RAM bank templates from the first FLASH (8000 - FFFF) Save download RAM bank templates is last FLASH (8000 - F2FF) Get download RAM bank templates from last FLASH (8000 - FFFF) Download/upload. Select/jump. Retrieve MSM6679A-110 firmware revision. Initialize background (BG) noise level. Retrieve vocabulary and trigger table revision number. Save SD templates from download RAM to first FLASH. Save SDR templates in last FLASH. (4A00-547BF300-FD7F) Recall SD templates from first FLASH to download RAM. Get SDR Templates from last FLASH (F300-FD7B4A00-547B) Select first FLASH bank for SI recognition. Set name tag length, set MSM665x busy mode ON. Set name tag record origin Set name tag record end Clear name tag table in SRAM (5480 - 56FF). Recall last saved name tag table. Recall name tag pointers from last FLASH (FD80-FFFF5480-56FF) Save name tag table from SRAM to FLASH. Save name tag pointers in last FLASH (5480-56FFFD80-FFFF) Set record volume high. Set record volume normal (default). Record name tag 01h - 3Dh. Set SD pointer to segment xxh. Search for SD utterance xxh. Enroll SD utterance selected by search command (F9xx). Erase utterance from SD vocabulary. Clear SDR table (4A00 - 547B)
Default (Hex) -- -- -- -- -- 3136 -- 3330 -- -- -- -- F509 0051 0000 01FF -- -- -- -- -- F50F F50F -- -- -- -- -- --
Request
Record
SD Recognition Control
F6xx F9xx FB00 FC00 F521
Response Summary
Command Operands F101h 00 tm F102h AdH AdL F103h AdH AdL F104h AdH AdL F11Xh F12Xh Description Record time = tm*14 msec. High and low bytes of SP/SI origin address. High and low bytes of SD origin address. High and low bytes of triggering origin address. IRQ Xh selected. SP table Xh selected. Invalid message received. Sample data over-run. [1] 32-Kbyte block boundary violation error. Unclassified download/upload error. Divide-by-zero error. Select/jump error. Invalid SP header or table. Reserved. Speech acknowledgment. [2]
Result after Parameter Set
F280h F240h F220h F210h Initialization Acknowledgment F208h F204h F202h F201h Speech Ack F400h
28
Semiconductor Response Summary (Continued)
Command F500h F501h F520h F540h F560h F580h F5A0h F5C0h F5E0h F5F0h F600h F6Utt F6 Utt Dst1H Dst1L...DstNH DstNL F6 Utt EminH EminL EmaxH EmaxL F6 Utt Dst1H Dst1L...DstNH DstNL EminH EminL EmaxH EmaxL F63Ah F63Bh F63Ch F63Dh F63Eh F63Fh F700h F73Eh F73Fh F740h F341h F7Utt SD Recognition F344h F7Utt DstH DstL F351h F7Utt Dst1H Dst1L... Result DstNH DstNL F361h F7Utt EminH EminL EmaxH EmaxL F371h F7Utt Dst1H Dst1L... DstNH DstNL EminH EminL EmaxH EmaxL Vector Upload Operands
MSM6679A-110 Voice Recognition Processor
Description MSM6679A-110 ready. Operation complete. Operations complete; MSM6679A-110 disabled (vocabulary 0). MSM6679A-110 waiting for start command. MSM6679A-110 waiting for end trigger. MSM6679A-110 processing recognition. Download/upload in progress. [4] Download/upload complete. Select/jump complete. Speak output in progress. Aborting SI listen mode. Utt = utterance ID. Utterance ID, high/low byte of distance to utterance 1...utterance N. Utterance ID, high/low byte of min. and max. energy value, Utterance ID, high/low byte of distance to utterance 1...utterance N, high/low byte of minimum energy value, high/low byte of maximum energy value. Trigger detection code (see init command). Rejection: utterance too loud. Rejection: utterance too long. Rejection: utterance begins too soon. Rejection: bad signal/noise ratio. Rejection: reason uncertain. Aborting SD Listen mode. After SD utterance search: not found. Rejection. Sort completed. After SD utterance search: empty. Rejection: MSM6679A-110 SD memory full/empty. After SD utterance search: in use. Utt = Utterance ID triggered. Utterance ID, high/low byte of distance. Utterance ID, high/low byte of distance to utterance 1... utterance N. Utterance ID, high/low byte of minimum energy value, maximum energy value. Utterance ID, high and low byte of distance to utterance 1... distance to utterance N, high and low byte of minimum energy value, maximum energy value.
Status [3]
SI Recognition Result [5]
F743h 0000h Upload failure. F743h NH NL V1H V1L...VNH VNL High/low bytes of length of vector, V, high/low byte of first V...Nth V. F801h F802h F804h F808h F810h F820h F840h F880h Reserved. Invalid SP header or table. Select/jump error. Divide-by-zero error. Unclassified download/upload error. Memory full; 32-Kbyte block boundary violation error. Sample data over-run. [1] Invalid message received. Record complete.
Trap Error Codes
Record Response FA00
29
MSM6679A-110 Voice Recognition Processor
Semiconductor
1.
2. 3. 4.
5.
Sample data overrun issued when real-time SP in Listen mode cannot keep up with incoming samples, i.e., if the A/D signal input routine overwrites a sample data buffer before it is fully processed. This acknowledge is sent only if Init command 1111 0010 xxxx x1xx (F2 xxxx x1xx) is set to enable acknowledgments. These messages are sent in response to a request command (F5XYh) from the host. Upload/download in progress, acknowledging load request immediately before data transfer. If in response to an N-byte download request, the MSM6679A-110 then receives N bytes (if N is even, or N+1 if N is odd) of data from the host. If N is odd and N+1 bytes are received, only N bytes are written to MSM6679A-110 memory. If in response to an upload, the MSM6679A-110 then sends N bytes (if N is even, or N+1 if N is odd) of data to the host. If an utterance was recognized, XYh is the utterance identity or class number, and additional parameters may be appended, if requested in the SI Recog (F3XYh with X=0...3) command. Otherwise, XYh indicates various results as detailed.
30
Semiconductor Command Descriptions Purge
Operand
MSM6679A-110 Voice Recognition Processor
Description Purge MSM6679A-110 Input Stack. This command clears the MSM6679A-110 input stack of commands that are waiting to be executed. Commands already in progress, such as a None pending MSM6654 poll action, are not affected. It does not affect the MSM6679A-110 output stack.
Return Values
F000
Set Parameter
Operand Description Set SP/SI Recognition Origin. Prior to SD or SI recognition, address pointers must be set to point at the SP or SI recognition parameter tables.This command sets the starting address of SP and SI recognition parameter tables. This address is the location of the first word of a header that contains pointers to one or more individual SP/SI tables. XXYYh = high (XXh) and low (YYh) bytes of requested address. The MSM6679A-110 uses and returns an even address outside the MSM6679A-110 work space that is as near as possible to the requested address. Leave this parameter at its default value unless you are using an Oki custom SI vocabulary and are instructed to alter SP/SI recognition origin. Default SP/SI origin: 8000h Return Values [1]
F102h XXYYh = High (XXh) and low (YYh) bytes of resultant address.
F102h XXYYh
If a valid header is not found at the resultant address, the MSM6679A-110 immediately sends response code: F802h = Invalid SP/SI header.
F103h XXYYh
Set SD Recognition Origin [2]. This command sets the SD origin address at the starting address of the current SD recognition parameter table. This command may be used to select among multiple RAM-resident SD vocabulary tables. XXYYh = high (XXh) and low (YYh) bytes of requested address. The MSM6679A-110 uses and returns an even F103h XXYYh = high (XXh) and address outside the MSM6679A-110 work space that is as low (YYh) of resultant address. near as possible to the requested address. Leave this parameter at its default value unless you are using an Oki custom vocabulary and are instructed to alter SD recognition origin. The table length is 0A7Ch bytes. Default SD origin: 4A00h Set Triggering Origin. This command sets the starting address of triggering parameter tables. This address is the location of the first word of a section of data memory containing one or more contiguous triggering parameter tables. XXYYh = high (XXh) and low (YYh) bytes of requested F104h XXYYh = high (XXh) and address. The MSM6679A-110 uses and returns an even low (YYh) bytes of resultant address outside the MSM6679A-110 work space that is as address. near as possible to the requested address. Leave this parameter at its default value unless you are using an Oki custom SI vocabulary and are instructed to alter triggering origin. Default triggering origin: F100h.
F104h XXYYh
31
MSM6679A-110 Voice Recognition Processor
Semiconductor
Set Parameter (Continued)
Operand Description Return Values [1]
F11Yh
Set IRQ Level. This command requests direction of host interrupts to IRQ Y. The MSM6679A-110 then selects IRQ Z, where Z is the nearest legal value to Y. Legal IRQ values are F11Zh = IRQ Z selected. any from the set {5 (default),A,B,C}. Default IRQ level: 5 Set SD Recognition SP table. This command sets the SP parameter table number to be used in processing speech input during SD Recognition. The MSM6679A-110 selects SP table number Z, where Z is the nearest valid value to Y. By default, the MSM6679A-110 selects SP table 3 until this command is issued. This command selects SP parameters only, and does not select among multiple RAM-resident SD vocabulary tables, which can be independently selected by the Set SD Origin command (F103h). After setting the table number and returning the resultant value, the MSM6679A-110 checks the validity of the SP header. If the header is invalid, an error message is returned. Set this value to (NSI +1), where NSI is the number of SI subvocabularies. Default SP table: 3. Select Triggering Table. This command selects triggering table TN for use with SP table VN. Valid values for VN and TN are between 01h and 0Fh. Leave this parameter at its default value unless you are using an Oki custom SI vocabulary and are instructed to alter the triggering table. Set ISA Mode. This command sets the port configuration for the ISA bus.
F12Z = SP table Z selected.
F12Yh
If the SP header is invalid, a second message follows: F802h = Invalid SP header.
F130h VN TN
F130h f(VN) f(TN) = Triggering table selected. Default = 0101, 0202, 0303...
F440h
None. Default is off.
1. 2.
Return value is actual parameter value which may not equal the set parameter value. See also F6XY
32
Semiconductor Initialize
F2xx Bit Power-On/ Values Reset Value
MSM6679A-110 Voice Recognition Processor
Action
Return Value
After power-on, the MSM6679A-110's mode corresponds to that after issuing a F20C command. This mode may NOT be the optimum condition for most situations, so the user is advised to carefully understand the desired condition and develop a suitable command for the application at hand. In addition, ensure that unwanted bits do not get set or reset when attempting to set individual conditions. The conditions selected are based on the XXh values associated with the last F2 command issued. Background Noise Initialization. When set to 1, the MSM6679A110 starts a 500-ms background noise initialization. When set to 0, the MSM6679A-110 does not perform background noise initialization. The MSM6679A-110 requires this command prior to recognition for noise vector subtraction during the utterance sampling period. Use the background initialization command whenever there is a change in the background noise level. For example, sample the noise signature in a vehicle at rest and moving at 35 MPH with its windows rolled down. The quality of a phone line connection can also vary from call to call. The host MCU must implement a strategy as to when to issue a background initialization command. In a vehicle, the host MCU could monitor the vehicle speed, fan speed, radio volume, etc. Alternatively, the host MCU could issue this command each time a new recognition session starts or a new line connection is established. However, the 0.5-sec sample period could degrade system responsiveness if used too frequently. A zero in this bit location during the F2XXh command will not cause an initialization. The F505h command causes the same initialization sequence. Wait for Recognition Command/Auto Restart SI Recognition. When set to 1, the MSM6679A-110 waits for a recognition command after each response. When set to 0, the MSM6679A110 auto-restarts SI recognition after each response. This bit should be set to 1 when an action is to be taken immediately after an utterance. Auto-restart recognition is the desired mode during digit string recognition, automated tape testing of digits, or in demonstrations where continuous recognition is desired.
F501 = Background initialization complete
1xxx xxxx
Cleared
F2XY = Initialization acknowledge. [1]
x1xx xxxx
Cleared
F2XY = Initialization acknowledge. [1]
33
MSM6679A-110 Voice Recognition Processor
Semiconductor
Initialize (Continued)
F2xx Bit Power-On/ Values Reset Value Action Return Value
xx1x xxxx
Cleared
Beep After Each Voice Trigger. When set to 1, the MSM6679A-110 beeps after each voice trigger. When set to 0, the MSM6679A-110 does not beep after each voice trigger. These beeps do not cause a F400h message to be issued to the host MCU. When set to 1, the MSM6679A-110 beep can help a user avoid speaking before the MSM6679A-110 is ready. This mode is normally used with a digits vocabulary to pace the user and confirm each utterance reception. F2XY = Initialization Instead of using beeps, an external MSM665x speech synthesizer acknowledge. [1] can repeat digits as they are recognized. However, some users find the number repetition annoying. Therefore, firmware could repeat digits during initial usage and switch to beep mode later. Typically, performance improves with time as users learns to speak with the correct enunciation and volumes. The MSM6679A-110 in this case trains the user. Note that the host MCU can also make the MSM6679A-110 beep with the F47Eh command. Set Output Volume. When set to 1, VOICEOUT1 sound output level is set to half of full volume (80h). When set to 0, voice output level is unaffected. MSM6679A-110 sound output volume can also be set at any level F2XY = Initialization on a continuous scale from 00h to FEh (low to high) with the acknowledge. [1] FEXXh command. The MSM665x speech synthesizer has four discrete sound output volumes, corresponding to 0h - 20h, 21h 40h, 41h - 80h, and 81h - FEh. Send Response Code After Sound Output. When set to 1, the MSM6679A-110 issues an acknowledge response (F400h) when sound output is completed. When set to 0, the MSM6679A-110 F2XY = Initialization does not issue an acknowledge response when speech response is acknowledge. [1] completed. Automatic beeps after voice triggers do not cause an F400h command to be issued. Trigger Detection Only. When set to 1, the MSM6679A-110 does not sort SI vocabularies for the best match, instead returning F63Ah code when an utterance has been detected. When set to 0, normal recognition is performed. When this bit is set to 1, the host MCU can use the F343h command to upload the recognition parameter vector, so that the host can perform independent processing. Clear SD Recognition and Name Tag RAM. When set to 1, the MSM6679A-110 initializes the SD parameter table. When set to 0, existing SD parameters are preserved. After this bit is set to 1, all SD training and name tag pointers are erased. Use this command to start training for a new user. If the old name tags are to be retained, the F50Ch command can recall old name tags from FLASH. To set up for a blank SD and name tag table at the next power-on, issue the command sequence F201h F507h.
xxxx 1xxx
Set
xxxx x1xx
Set
xxxx xx1x
Cleared
F2XY = Initialization acknowledge. [1]
xxxx xxx1
Cleared
F2XY = Initialization acknowledge. [1]
1.
See the Response Summary table earlier in this section for a complete description of the XY codes in initialization acknowledgment messages.
34
Semiconductor Recognize
Opcode F300h Action
MSM6679A-110 Voice Recognition Processor
Return Value None F600h F700h F600h MSM6679A-110 was not in Listen mode. Aborting SI Listen mode. Aborting SD Listen mode. Aborting SI Listen mode. Trigger detection code (see Initialization command). Invalid signal processing table. Sample data overrun. Utterance ID in vocabulary Y.
Stop Listening. This command causes the MSM6679A-110 to exit SI or SD Listen mode, whichever was active. Start SI Listen Mode. For all the following opcodes, the MSM6679A-110 performs SI recognition on incoming utterances, using SI vocabulary Y. The vocabulary Y is identified by one of 15 sets, thus Y = 1h ~ Fh. Return recognized phrase using vocabulary number Y. Return recognized phrase and distance table for vocab Y. Return recognized phrase and energy value for vocab Y.
F63Ah
F63Bh~F63Fh Rejection. F802h F840h
F30Yh F301h F33Fh F31Yh
F6h Utt
Utterance ID in vocabulary Y, high and F6h Utt Dst1H Dst1L... low byte of distance to utterance 1... DstNH DstNL distance to utterance N. Utterance ID in vocabulary Y, high and F6h Utt EminH EminL low byte of minimum and maximum EmaxH EmaxL energy value. F6h Utt Dst1H Dst1L... DstNH DstNL EminH EminL EmaxH EmaxL F740 F700 F73E F73F F802 F840 Utterance ID, high and low byte of distance to utterance 1...distance to utterance N, high and low byte of minimum and maximum energy value. Triggered. Abort SD Listen mode. Rejection. Memory empty. Invalid SP table. Sample data overrun. Abnormal response: Memory empty.
F32Yh
F33Yh
Return recognized phrase, distance table, and energy value for vocab Y.
F340h
Start SD Listen Mode. When an utterance is captured, it is analyzed and converted to a "recognition parameter vector." The host may then command the MSM6679A-110 to use this vector in various ways (e.g., Sort, Update, or Recognition Vector Upload). SD Recognition Sort. These commands sort the distances between the recognition parameter vector and the reference vectors for the utterances in the current SD vocabulary. Return recognized phrase for vocab Y. This command can be issued several times to yield first, second, third best, etc. Return recognized phrase and distance for the current vocabulary. Return recognized phrase and distance table for vocab Y. Return recognized phrase and energy value for vocab Y.
F73Fh
F341h, F344h, F351h, F361h, F371h
F341h
F7h Utt
Utt= Utterance ID.
F344h
F7h Utt DstH DstL
Utt = index of recognized phrase, DstH DstL = high/low bytes of distance from nearest phrase.
F351
F7h Utt Utterance ID, high and low byte of Dst1H Dst1L... distance to utt. 1...N. DstNH DstNL F7h Utt Utterance ID, high and low byte of EminH EminL minimum and maximum energy value. EmaxH EmaxL
F361h
35
MSM6679A-110 Voice Recognition Processor
Semiconductor
Recognize (Continued)
Opcode F341h, F344h, F351h, F361h, F371h Action F7h Utt Dst1H Dst1L... DstNH DstNL EminH EminL EmaxH EmaxL Return Value Utterance ID, high and low byte of distance to utterance 1...distance to utterance N, high and low byte of minimum and maximum energy value.
F371h
Return recognized phrase, distance table, and energy value for vocab Y.
F342h
Update SD Recognition Enrollment. This command updates enrollment on utterance Utt, immediately after a "F7h Utt" response to the Sort SD Distances command (F341h). Alternatively, the utterance to be updated can be selected by the SD Search command (F9XYh). This command uses the recognition parameter vector from the most recently captured utterance, and does not start SD Listen mode. Generally, update should be performed only if correct utterance identify is confirmed by the user.
F740h
Update complete.
F343h
Recognition Vector Upload. Request recognition parameter vector upload to host.
F743h NH NL V1H V1L... VNH VNL = Success, where NH/NL = high/low bytes of N, N = Length of recognition parameter vector V, V1H/V1L = high/low bytes of first element of V, VNH/VNL = high/low bytes of Nth element. F743h 00 00 Failure.
Speak
Opcode Action Return Value
Speak Phrase from External Memory. This command causes the MSM6679A-110 to play back a name tag from external memory. If no F401h ~ sound is defined for a selected index, the F43Dh MSM6679A-110 plays a beep. See the Record commands for information on creating name tags. Speak Phrase from Low Internal Memory. If no sound is defined for a selected index, the MSM6679A-110 plays a beep. The default phrases supplied with the MSM6679A-110 in the smaller low playback memory area are F441h ~ listed below. F450h F441h Drip. F442h F443h F444h Buzzer. Dial tone. Bonk.
F400h
If enabled, this value is returned upon completion of playback.
F400h
If enabled, this value is returned upon completion of playback.
36
Semiconductor
MSM6679A-110 Voice Recognition Processor
Speak (Continued)
Opcode Action Speak Phrase from High Internal/External Memory. If no sound is defined for a selected index, the MSM6679A-110 plays a beep. The default phrases supplied with the MSM6679A110 in the larger upper playback memory area are listed below. F451h F452h F453h F451h ~ F454h F47Ch F455h F456h F457h F458h F459h F45Ah F45Bh F45Ch F47D F47Eh "0" simulated DTMF tone. "1" simulated DTMF tone. "2" simulated DTMF tone. "3" simulated DTMF tone. "4" simulated DTMF tone. "5" simulated DTMF tone. "6" simulated DTMF tone. "7" simulated DTMF tone. "8" simulated DTMF tone. "9" simulated DTMF tone. "*" simulated DTMF tone. "#" simulated DTMF tone. -- F400h -- If enabled, this value is returned upon completion of playback. F400h If enabled, this value is returned upon completion of playback. Return Value
Reserved. This command is reserved for future use. Beep. This causes the MSM6679A-110 to beep for 50 ms.
F47Fh
Pause. This command can be issued while the MSM6679A-110 is performing sound output and is then put in the MSM6679A-110 command stack for subsequent processing. F400h When this command is executed, sound output pauses for 0.2 sec. The pause command is useful for word spacing. Set MSM6654 Mode. This command causes the MSM6679A-110 to initialize None. the external MSM665x device, also clearing the device from BUSY mode.
If enabled, this value is returned upon completion of playback.
F480h
Playback Sound from MSM665x Device. This command causes the MSM6679A-110 to issue a speak command to the MSM665x slave device. F481h The value is passed on the MSM665x device as F400h F4FFh 01h - 07Fh. The actual phrase is determined by the vocabulary programmed into the MSM665x device. Up to 127 external phrases are supported. F50Bh Set MSM665x Busy Mode ON. None.
If enabled, this value is returned upon completion of playback. If NAR is set, the F400h command is sent when the MSM665x device is ready for another command. If busy mode is selected, the F400 command is returened when the sound is finished.
37
MSM6679A-110 Voice Recognition Processor
Semiconductor
Speak (Continued)
Opcode Action Set 6654 NAR mode. This command, which is the complement of the F50B command, sets up the handshaking to the attached 6654 speech None. synthesizer to use the NAR. This setup uses the 6654's double buffer feature to eliminate any gap between two consecutive phases. Set Output Level. This command sets the speech output level to one of 255 values as follows: FEXYh FE03 FE80h FEFEh Set minimum output level. Set output level half way (default). Set maximum output level. None. Return Value
F51Bh
Request
Opcode Action F500h F520h F540h F500h Status Request. This command causes the MSM6679A-110 to return a 2-byte value indicating its current status. F560h F580h F5A0h F5C0h F5E0h F501h Select last FLASH bank for SI recognition. Select download RAM bank for SI/SP template area. This command enables the download RAM bank in the upper 32 K of data memory for SI recognition. Select buffer RAM bank for SI/SP. This command enables the buffer RAM bank in the upper 32 K of data memory for SI recognition. Return Value MSM6679A-110 ready. MSM6679A-110 disabled. MSM6679A-110 waiting for start. MSM6679A-110 waiting for end. MSM6679A-110 processing. Download/upload in progress. Download/upload complete. Select/jump complete.
F510h
No return value
F520h
No return value
F522h
Copy download RAM bank to buffer RAM bank. This command copies the download RAM bank F501h to the buffer RAM bank. The copied address range is (8000-FFFF). Save download RAM bank templates in first FLASH. Save the download RAM SI/SP area (8000-F2FF) to the same address range in the first FLASH.
Copy is complete.
F513h
F501h
Save is complete.
38
Semiconductor
MSM6679A-110 Voice Recognition Processor
Request (Continued)
Opcode Action Get download RAM bank templates from the first FLASH. Recall the download RAM SI/SP template (8000 - FFFF) from the same address range in the first FLASH. Save download RAM bank templates in last FLASH. SAVE the download RAM bank SI/SP template area (8000 - F2FF) to the same address range in the last FLASH. Get download RAM bank templates from last FLASH. Recall the download RAM bank SI/SP template area (8000 - FFFF) to the same address range in the last FLASH. Download/Upload.
Full syntax: F5 02 00 Ctl AdH AdL NH NL [Dt1... DtN [Dt(N+1)]] Full syntax: F5 02 00 Ctl AdH AdL NH NL [Dt1... DtN [Dt(N+1)]] Ctl(7) = 0 for download, Ctl(7) = 1 for upload Ctl(6) = 0 for data RAM, Ctl(6) = 1 for program RAM/ROM If Ctl(6)=0 then Ctl(1-0) = Seg: Data segment selection If Ctl(6)=1 and Ctl(1-0) = x0, then external program segment 0 is used. If Ctl(6)=1 and Ctl(1-0) = x1, then external program segment 1 is used.
Return Value
F514h
F501h
Save is complete.
F515h
F501h
Save is complete.
F516h
F501h
Save is complete.
Immediately after receiving parameter NL, the MSM6679A-110 responds with a message to indicate acceptance or denial of the transfer request. Acceptance is indicated by F5A0h. Denial is indicated by a F8XYh. At the end of an accepted transfer, the MSM6679A-110 responds with a message to confirm or deny valid completion of the transfer. Valid completion is indicated by F5C0h.
F502h
AdH AdL = high, low bytes of starting address. NH NL = high, low bytes of N N = Number of bytes to be downloaded or uploaded (maximum 07FFCh) Dt1... DtN = Download data. Note (here and in upload response) that data are 8-bit binary values, even if using the serial interface. Dt(N+1). If N is odd, an extra byte is appended to the data so that the total number of bytes in the message remains even. This command requests data transfer to/from data or external program memory.The control parameter (Ctl) controls the direction of the transfer (i.e., download vs. upload) and specifies which of six 64-Kbyte memory segments (i.e., four data segments and two external program segments) is to be accessed. This command does not work with internal program memory. It is not possible to download to external program memory while running in external program memory. The address and length parameters (AdH AdL NH NL) specify the starting address and length of the transfer in bytes. Since the MSM6679A-110 can only perform download /upload transfers within one 32-Kbyte block in one Download /Upload command, the address and length parameters must not specify a transfer that violates a 32-Kbyte address boundary. If this restriction is violated, the download/upload request will be denied.
F880h F840h F820h F810h F808h F804h F802h F801h FAXYh FBXYh
Invalid message received. Sample data over-run. 32-Kbyte block boundary violation error. Unclassified download/upload error. Divide-by-zero error. Select/jump error. Invalid SP header or table. Reserved. Most and least significant byte of address where error occurred.
39
MSM6679A-110 Voice Recognition Processor
Semiconductor
Request (Continued)
Opcode Action Return Value
Select/Jump. This command selects a new data segment, or Jumps to a new program segment. Ctl(7)=0 is used to first select a new data segment. Ctl(7)=1 then jumps to that program segment. Upper 32-Kbyte of Seg(7)=0 selected segment is accessed normally. Access lower 32-Kbyte Ctl(7)=0 Seg(7)=1 block of selected segment in upper 32 Kbytes of data space. Seg(6~2) Reserved. F503h Ctl Seg Seg(1~0) Data segment selection. Seg(7)=0 Jump to selected external program segment. F5E0h Success. F5E0h Success.
F8XYh
Failure, with XY(2) = 1.
Seg(7)=1 Jump to internal program segment. Seg(6~1) Reserved. Ctl(7)=1 If Seg(7) =1, not used. If Seg(7) = 0 and Seg(0) = 0: external program segment 0. If Seg(7) = 0 and Seg(0) = 1: external program segment 1.
Seg(0)
F8XYh
Failure, with XY(2) = 1.
F504h
Retrieve MSM6679A-110 Firmware Revision Number.
XXXX
Four-digit ASCII number.
40
Semiconductor Request (Continued)
Opcode Action
MSM6679A-110 Voice Recognition Processor
Return Value
F505h
Initialize in Background. Background noise initialization is performed for 500 ms. The MSM6679A-110 requires this command prior to recognition for noise vector subtraction during the utterance sampling period. Use the background initialization command whenever there is a change in the background noise level. For example, sample the noise signature in a vehicle at rest and moving at 35 MPH with its windows rolled down. The quality of a phone line connection can also vary from call to call. The host MCU must implement a strategy as to when to issue a background initialization F501h command. In a vehicle, the host MCU could monitor the vehicle speed, fan speed, radio volume, etc. Alternatively, the host MCU could issue this command each time a new recognition session starts or a new line connection is established. However, the 0.5-sec sample period could degrade system responsiveness if used too frequently. A zero in this bit location during the F2XXh command will not cause an initialization. The F2xxh command can also be used to perform background noise initialization. Retrieve Vocabulary and Trigger Table Revision XXXX Number. Save SDR templates in last FLASH. Save the download RAM bank SD template area. Saves 2684 bytes from the address set by the F103 command to the address range F300FD7F in the last FLASH. The default is 4A00547BF300-FD7F). Get SDR templates from last FLASH. Get the download RAM bank SD template area. Saves 2684 bytes to the address set by the F103 command from the address range F300FD7B in the last FLASH. The default is (F300FD7B4A00-547B). Select Default SI Vocabulary. (First FLASH)
Initialization is complete.
F506h
Four digit ASCII number.
F507h
F501h
Save is complete.
F508h
No return value
F509h
--
--
41
MSM6679A-110 Voice Recognition Processor
Semiconductor
Record
Opcode Action Set Name Tag Length, Set MSM665x Busy Mode ON. Name tag record length is set by XXh, with XXh defining record length in 14-ms intervals. The maximum record length of FFh yields a recording interval of 3.57 sec. The default value is 1.2 sec. Set Name Tag Record Origin. This command sets the beginning address for recording name tags. XXXX = 128 byte blocks from 0000 to 02FF. The reset default is 0000. This is only effective before an F50A command since new recordings start after the end of the previous recording. The F50A command uses this number to calculate the first address. Return Value
F101h 00XXh
F101h 00XXh Operation complete.
F105 xxxx
F105 BAAA, where B is the bank number (0,1,2), and AAA is the bank address /16 (800 - FF8)
F106 xxxx
F106 BAAA, where B is the Set Name Tag Record End. This command sets bank number the ending address for recording name tags. (0,1,2), and XXXX = 128 byte blocks from 0000 to 02FF. AAA is the The reset default is 01FF. bank address /16 (800 - FF8) Clear Name Tag Table. F501h Name tag table cleared. Recall name tag pointers from first FLASH. Save the first FLASH name tag pointers (FD80 F501h FFFF) to the working name tag pointer table. The default is (FD80-FFFF5480-56FF). Recall name tag pointers from last FLASH. Save the last FLASH name tag pointesr (FD80 FFFF) to the working name tag pointer table. The default is (FD80-FFFF5480-56FF). Save name tag pointers in first FLASH. Save the working name tag pointer table to the first FLASH name tag pointers. The default is (5480 -56FDFD80-FFFD).
F50Ah
F50Ch
Saved name tag table recalled.
F51Ch
F501h
Name tag pointers recalled.
F50Dh
F501h
Name tag table saved.
F51Dh
Save name tag pointers in last FLASH. Save the working name tag pointer table to the last F501h FLASH name tag pointers. The default is (5480 -56FDFD80-FFFD). Set Record Volume HIGH. Set Record Volume to Normal. This is the default setting. Reserved. This command is reserved for future use. -- -- -- FA00h F280h
Name tag pointers saved.
F50Eh F50Fh FA00h
-- -- -- Completed. Memory full.
FA01h ~ FA3Dh Record Name Tag.
42
Semiconductor
MSM6679A-110 Voice Recognition Processor
Record (Continued)
Opcode Action -- -- Return Value
FA3Dh ~ Reserved. These commands are reserved for FAFFh future use.
SD Recognition Control
Opcode Action Return Value
Recognition performance is largely a function of how well the enrollment data represents subsequent tokens of the enrolled utterances, and performance generally improves steadily with each additional enrollment pass. For most applications, three initial enrollment passes are recommended. Subsequent reference updating can be performed with the SD Recognize Update command (F342). Clear SDR table. This command initializes a blank SD template table. The 2684-byte area from the address set by the F103 command (the working SDR table) is set to zeros. The SDR tables in the FLASH banks are not affected. The default is (4A00 - 547B). Set SD Segment Pointer. This command sets the SD segment pointer to XY00h, i.e., set the starting address of the current SD recognition parameter table to XY00h. Issuing this command is equivalent to issuing the Set SD Origin command, F103h XY00h. (For further details of operation, please refer to the description of that command.) Search for SD Utterance XY. This is the first step in adding an utterance to the vocabulary, or in replacing an existing one. The SD vocabulary memory is searched for utt. no. XYh. If it is not found and if sufficient SD memory exists, the MSM6679A prepares to add utterance number XYh to the vocabulary. Enroll SD Utterance. This command starts MSM6679A SD Listen mode, then uses the next captured utterance to start or update training of the reference data for SD utterance number XY specified in the most recent Search command (F9XYh). The user must be prompted to say the utterance prior to issuing this command. If the utterance was previously enrolled, a training update is performed; if not, the reference data is initialized. Each utterance in the SD vocabulary must be enrolled at least once before it can be recognized.
F521h
F501h
SDR table is cleared
F6XYh
No return value.
F740h F700h F73Fh F740h F700h F73Eh F802h
Utterance number found. Utterance number not found. Memory full. Operation complete. Aborting SD Listen mode. Improper level, must repeat. Invalid signal processing table.
F9XYh
FB00h
F840h
Sample data overrun.
FC00h
Erase utterance from SD vocabulary. This command erases the reference parameters for utterance number XYh from the SD vocabulary, F740h where XYh is the utterance number retained from the previous Search command (F9XYh).
Operation complete.
43
MSM6679A-110 Voice Recognition Processor
Semiconductor
Asynchronous Serial Protocol Example All messages to the MSM6679A (except downloads and uploads) are echoed, but replies from the MSM6679A to the host are not echoed by the host. This arrangement facilitates manual communication with the MSM6679A using standard terminals. The following table illustrates the range of MSM6679A functions.
Comment Initialize MSM6679A Install new software ersion. Action Host initializes MSM6679A. MSM6679A acknowledges. Host requests download to program segment 40, starting at location 0, of 32 Kbytes (7FFCh). MSM6679A accepts request. Host sends 32 Kbytes. (~34 sec at 9600 baud). MSM6679A indicates downloadcomplete. Host requests upload from program segment 0, starting at location 0, of 32 Kbytes (7FFCh). MSM6679A accepts request. MSM6679A sends 32 Kbytes. MSM6679A indicates upload complete. Host commands jump to external program segment 0. MSM6679A begins running new load. Host requests download to data segment 0, starting at location 5000h, of 256 bytes (0100h). MSM6679A accepts request. Host sends 256 bytes (~0.25 sec at 9600 baud). MSM6679A indicates download complete. Host requests Set triggering origin to 5000h. MSM6679A sets triggering origin and sends confirming response. Host requests download to data segment 0, starting at location 6000h, of 4 Kbytes (1000h). MSM6679A accepts request. Host sends 4 Kbytes (~4.3 sec at 9600 baud) MSM6679A indicates download complete. Voice Input Host F258 F502 0040 0000 7FFC F5A0 ... MSM6679A F258 F200 502 0040 0000 7FFC
Command Response
F5C0 F502 00C0 0000 7FFC F502 00C0 0000 7FFC F5A0 ... F5C0 F503 8000 F5E0 F502 0000 5000 0100 F5A0
Upload software for verification of transfer.
Run new software.
F503 8000 F502 0000 5000 0100 ...
Load trigger tables at 5000h.
F5C0 F104 5000 F104 5000 F104 5000 F502 0000 6000 1000 F5A0
Set new triggering origin.
Download new SD vocabulary.
F502 0000 6000 1000 ...
F5C0
44
Semiconductor
MSM6679A-110 Voice Recognition Processor
Comment Set new SD tables.
Action Host requests Set SD origin to 6000h. MSM6679A sets SD origin and responds. Host requests download to data segment 0, starting at location 7000h, of 4k bytes (1000h). MSM6679A accepts request. Host sends 4 Kbytes. MSM6679A indicates download complete. Host requests download to data segment 0, starting at location 8000h, of 32k bytes (7FFC). MSM6679A accepts request HOST sends 32 Kbytes. MSM6679A indicates download complete. Host requests Set SP/SI origin = 7000h. MSM6679A sets SP/SI origin and responds. Host requests upload from data segment 0, starting at location 300h, of 45 bytes (2Dh). MSM6679A accepts request, signals in progress. MSM6679A sends 46 bytes. MSM6679A indicates upload complete. Host requests set SP table 3. MSM6679A selects SP table 3 and confirms. Host initializes MSM6679A. MSM6679A acknowledges. Host starts SI recognition, vocabulary 1.
Voice Input
Host F103 6000
MSM6679A F103 6000 F103 6000 F502 0000 7000 1000 F5A0 F5C0
Command Response
Download first 4 K of SI vocabulary.
F502 0000 7000 1000 ... F502 0000 8000 7FFC ...
Download last 32 K of SI vocabulary.
F502 0000 8000 7FFC F5A0 F5C0
Set new SP/SI tables.
F102 7000
F102 7000 F102 7000 F502 00A0 0300 002D F5A0 ... F5C0
Upload data for diagnostics.
F502 00A0 0300 002D
Set up MSM6679A for SI recognition.
F123
F123 F123 F258 F200 F301 F603 F302 F602 F302 F603
F258 F301 "Dial"
SI recognition.
MSM6679A recognizes utterance 3. Host starts SI recognition, vocabulary 2. "Two" MSM6679A recognizes utterance 2. Host starts SI recognition, vocabulary 2. "Three" MSM6679A recognizes utterance 3.
F302
F302
45
MSM6679A-110 Voice Recognition Processor
Semiconductor
Comment SD enrollment.
Action Host starts SI recognition, vocabulary 1.
Voice Input
Host F301
MSM6679A F301 F607 F901 F700 FB00 F740 FB00 F740 FB00 F740
Command Response
"Store" MSM6679A recognizes utterance 7. Get ready to train SD utterance 1. Memory is empty and ready to train. Pass 1; host sends SD enroll command. "John Smith" SD utterance 1 initialized. Pass 2; host sends SD enroll command. "John Smith" SD utterance 1 updated. Pass 3. Host sends SD enroll command. "John Smith" SD utterance 1 updated. SI recognition of control words. Host starts SI recognition, vocabulary 1. "Dial" MSM6679A recognizes utterance 3. Host starts SI recognition, vocabulary 2. "Five" MSM6679A recognizes utterance 5. Host starts SI recognition, vocabulary 2. "Six" MSM6679A recognizes utterance 6. Host starts SI recognition, vocabulary 1. "Store" MSM6679A recognizes utterance 7. SD enrollment. Host prepares MSM6679A to train SD utterance 2 Memory is empty and ready to train. Pass 1; host sends SD enroll command. "Bill Jones" SD utterance 2 initialized. Pass 2; host sends SD enroll command. "Bill Jones" MSM6679A updates SD utterance 2. Pass 3; host sends SD enroll command. "Bill Jones" MSM6679A signals operation completed. SI recognition of control word. SD recognition. Host starts SI recognition, vocabulary 1. "Call" MSM6679A recognizes utterance 11. Host starts SD recognition. "John Smith" MSM6679A signals trigger OK. Host sends SD sort command. MSM6679A recognizes utterance 1. F341 F740 F341 F701 F340 F60B F340 F301 F740 F301 FB00 F740 FB00 FB00 F740 FB00 F902 F607 F902 F700 FB00 F301 F606 F301 F302 F605 F302 F302 F603 F302 F301 F301 FB00 FB00 F901 FB00
FB00
46
Semiconductor
MSM6679A-110 Voice Recognition Processor
Comment Name tag recording.
Action Host initiates MSM665x port. Host sets recording length to 1 sec. MSM6679A signals operation complete. Host clears name tag table MSM6679A signals operation complete. Host sets record gain to max. level. Start recording tag one.
Voice Input
Host F480 F101 0047 F50A F50E FA01
MSM6679A F480 F101 0047 F101 0047 F50A F501 F50E FA01 FA00
Command Response
"Jane Doe" MSM6679A signals name tag recording complete. Save name tags to FLASH. Name tags saved. Name tag playback. Host sets volume to max. level. Host commands play back name tag 1. MSM6679A signals playback OK. Sound playback. Host sets output volume to mid point. Play MSM6679A internal sound 1. Play back sound from MSM6654. FE80 F442 F49F
F50D FEFF F401
F50D F501 FEFF F401 "Jane Doe" F400 FE80 F442 "bzzzz" F49F "Completed"
The information contained herein can change without notice owing to product and/or technical improvements. Please make sure before using the product that the information you are referring to is up-to-date. The outline of action and examples of application circuits described herein have been chosen as an explanation of the standard action and performance of the product. When you actually plan to use the product, please ensure that the outside conditions are reflected in the actual circuit and assembly designs. OKI assumes no responsibility or liability whatsoever for any failure or unusual or unexpected operation resulting from misuse, neglect, improper installation, repair, alteration or accident, improper handling, or unusual physical or electrical stress including, but not limited to, exposure to parameters outside the specified maximum ratings or operation outside the specified operating range. Neither indemnity against nor license of a third party's industrial and intellectual property right,etc.is granted by us in connection with the use of product and/or the information and drawings contained herein. No responsibility is assumed by us for any infringement of a third party's right which may result from the use thereof. When designing your product, please use our product below the specified maximum ratings and within the specified operating ranges, including but not limited to operating voltage, power dissipation, and operating temperature. The products listed in this document are intended for use in general electronics equipment for commercial applications (e.g.,office automation, communication equipment, measurement equipment, consumer electronics, etc.).These products are not authorized for use in any system or application that requires special or enhanced quality and reliability characteristics nor in any system or application where the failure of such system or application may result in the loss or damage of property or death or injury to humans. Such applications include, but are not limited to: traffic control, automotive, safety, aerospace, nuclear power control, and medical, including life support and maintenance. 47
MSM6679A-110 Voice Recognition Processor
Semiconductor
Certain parts in this document may need governmental approval before they can be exported to certain countries. The purchaser assumes the responsibility of determining the legality of export of these parts and will take appropriate and necessary steps, at their own expense, for export to another country. Copyright 1997 OKI SEMICONDUCTOR OKI Semiconductor reserves the right to make changes in specifications at anytime and without notice. This information furnished by OKI Semiconductor in this publication is believed to be accurate and reliable. However, no responsibility is assumed by OKI Semiconductor for its use; nor for any infringements of patents or other rights of third parties resulting from its use. No license is granted under any patents or patent rights of OKI.
48

▲Up To Search▲

Price & Availability of MSM66P56

	To Download MSM66P56 Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .