circuitcellar.com
Magazine Support   Digital Library   Products & Services   Suppliers Directory 
 
 





 

September 1999, Issue 110

Taking Orders:
A Speech Recognition Module


by John Iovine

VoiceDirect PROCESSOR

The heart of this module is the VoiceDirect speech recognition processor. The processor is available in a QFP-64 package for anyone who needs or wants to build a circuit from scratch.

This module has a lot of the preliminary work already provided, including an AGC audio amplifier, serial EEPROM, and clock. It also contains extensive socket headers (JP1, JP2, and JP3), which make it easy to connect an external circuit to the module. Table 1 shows you the pinout.

Name Module Pin Description
  JP1 1-17 Unused
  JP2 9 Unused
  JP3 1-9 Unused
Preamp in JP2 1 Microphone input connection
Mic Bias JP2 2 Mic bias (electret microphone)
AGND JP2 3,5 Analog ground–do not connect to digital ground because of noise
+5V JP2 4 Vcc
PWM1 JP2 6 Pulse width modulator–Output 1, connects to 8–32-Wspeaker
PWM0 JP2 7 Pulse width modulator–Output 0, connects to 8–32-Wspeaker
DACOUT JP2 8 Analog audio ouput–Provides better quality sound than PWM ouput, requires amplifier
RECOG JP2 10 Recognition sensitivity selection and activates recognition
TRAIN JP2 11 Training sensitivity selection and activates training
Out1-Out7 JP2 12-18 Stand-alone mode output lines 1-7
High/Out8 JP2 19 Stand-alone mode output lines 8/ or high
ERROR JP3 10 Stand-alone mode error signal
GND JP3 11,12 Digital ground, CPU core (pins 1, 33)
Mode JP3 13 Stand alone or slave
Reset JP3 14 Vcc

Table 1—Here you see the pinout for the module. The schematic is given in Figure 1.

We can best explore the capabilities of the module by getting it up and running. The VoiceDirect speech-recognition kit contains the module (assembled), microphone, speaker, three microswitches, two 100-kW resistors, and a quick setup guide (see Photo 2).

9906017-p2.jpg (100946 bytes)

Photo1—The kit includes the module, resistors, three push-button switches, microphone, speaker, and manual—all the external components you need to implement stand-alone recognition.

The module’s schematic is shown in Figure 1. The few external components that are supplied with the kit are all it takes to get the module functioning. The PCB measures 2² × 2² and has 0.1² header sockets soldered to one side of the board, which makes it easy to connect to the circuit. To experiment with the module, I placed my external components on a solderless breadboard (see Figure 2).

9906017fig2.gif (19614 bytes)

Figure 2—This test circuit schematic demonstrates how simple it is to get the VoiceDirect module up and running.

To connect components on the breadboard to the socket headers, I used 22-AWG stranded wire. I mounted the microswitches, resistors, microphone, and LEDs on the breadboard.

One note here: the schematics in the manual detail the board shown from the top, but the header sockets are mounted on the bottom side of the PCB. This is in contrast to Figure 2, which shows the board with the header sockets on top. Be careful when comparing Figure 2 to the drawing in the manual.

The VoiceDirect module recognizes 15 words in stand-alone mode and has only eight outputs (connector JP2 pins 12–19). For simplicity, I’m only using eight outputs (eight words) so I don’t need to add a decoding circuit. The 8-pin output for the 15-word recognition doesn’t follow the standard binary numbering (see Table 2).

Word 1 Output 1
Word 2 Output 2
Word 3 Output 3
Word 4 Output 4
Word 5 Output 5
Word 6 Output 6
Word 7 Output 7
Word 8 Output 8
Word 9 Output 8 and 1
Word 10 Output 8 and 2
Word 11 Output 8 and 3
Word 12 Output 8 and 4
Word 13 Output 8 and 5
Word 14 Output 8 and 6
Word 15 Output 8 and 7

Table 2—This table shows the digital output for 15 words when the device is in stand-alone mode.