Results

The video describes the FVE.

Coming to the results, we have been able to satisfy all the listed performance requirements and in some cases have performed even better than we expected.

Voice Command Test Results:

Performed tests for 50 iterations in noiseless and noisy environments to check the accuracy of the recognition system.
Tested for:
– Recognizing the words and performing suitable action
– Recognizing incorrect word
– Rejecting Garbage before activation
– Rejecting Garbage after activation

Results in Noiseless Environment:
• In a noiseless environment, the accuracy of recognizing the activation word is 86%
• The accuracy of recognizing Alpha, Bravo and Charlie is 76%, 92%, and 90% respectively
• The overall mean accuracy is 86%
• Garbage words before “Computer” are completely rejected with an accuracy of 100%
• Garbage words after “Computer” are all accepted with 0% correct results (none are rejected)

Results in Noisy Environment:
• In a noisy environment, the accuracy of recognition goes down drastically
• The accuracy in recognizing the activation word “Computer” is 50%
• The accuracy of recognizing Alpha, Bravo and Charlie is 58%, 74%, and 68% respectively
• The overall mean accuracy is 62.5%
• Garbage words before “Computer” are completely rejected with an accuracy of 100% in static noise, all bets are off in dynamic noise
• Garbage words after “Computer” are all accepted with 0% correct results (none are rejected)