Vehicle Voice Control System
by Voice Template Matching
Submitted by: Ed Leslie, Pyi Lwin and Todd Gardner
Submitted to: Dr. Winncy Du
Department of Mechanical & Aerospace Engineering
San Jose State University
May 20, 2009
This initial proof of concept project was intended to be the first phase (~4 weeks) ofa Vehicle Voice Control System (VVCS). During the course of this project, methods of voice recognition would be explored, tested, and the operation would be verified in conjunction with a microcontroller, which controls hardware. Areas of improvement would then be identified with the intent to address them in some future project.
a. Voicetemplate “matching” was moderately successful and indicated correct and incorrect “matches” with software and hardware signals.
i. “Match” indication: After some experimentation, the “match” range was set to be within 2% of the minimum standard deviation of the difference between the power spectral density (PSD) of the template and the PSD of the“psuedo-real-time” signal was used to indicate a “match”.
ii. Template creation: The acquisition, synchronization and averaging is achieved to create a voice templates.
iii. Testing: The system was tested with different amounts of white and structured noise, time compression/expansion and delay distortion to the acquired voice templates.
iv.“Psuedo-real-time”: Level detection was performed on the incoming “real-time” audio stream to start a short acquisition of 216 samples at 44.1 KHz (It is a coincidence that 216 levels may exist in the DAC acquisition hardware). This yielded operation that seemed to be in real-time.
i. RS-232 commands: Commands were transmitted from the LabVIEW computer to the microcontrollervia the RS-232 serial port.
ii. Hardware timers: Hardware timers were used instead of delays to time the switching of hardware components. This helped to avoid delay functions, which cause idle time during with the microcontroller does nothing useful.
iii. Interrupts: To allow multiple processes to run at the same time, interrupts were used. An interrupt isgenerated when a command is received via RS-232 or when a hardware timer reaches a time limit at which an action needs to be taken.
a. Voice Template “Matching”: Four physical components were activated by voice:
i. Right signal (LED)
ii. Left signal (LED)
iii. Hazard flashers (LEDs)
iv. Horn (electronic beeper)
b.Error display: An error in command recognition is indicated by an LED if any of the following occur:
i. Templates with different numerical prefixes fall within the match range.
ii. A “match” occurs to a template having a numerical prefix of 00.
iii. A match occurs to a template that was not assigned a numerical prefix.
c. A “no match” condition isnot possible because there is always a minimum standard deviation when comparing two numbers.
Imagine yourself every day having to rely on somebody else to take you everywhere you need to go. If you need to take your spouse to the doctor, buy groceries, go to your child's school play or just go to work you need to wait. You must plan to perfection days in advance because makingchanges to the course of action that best drives your life may cause irreversible failures in your life.
There are many people that could benefit from a non-critical vehicle voice control system (VVCS). Able-bodied people, and those that need to be “REABLED” that drive could also benefit by being able to control various systems by voice. There are currently only a handful of controls...