Voice Recognition is used to recognize a speaker's voice through his/her pitch and Frequency. We aim to recognize a speaker at real-time. We use two modules – feature extraction and codebook generation. We have extracted voice features by computing Mel Frequency Cepstrum Coefficients. Speaker specific codebook is generated by the Vector Quantization technique introduced by Linde, Buzo and Gray. In the training phase, speaker specific codebook is computed and stored in the database against each speaker. At the time of recognition, input voice signal is recorded and its codebook is computed and compared with the codebooks stored in the database. If a match is found the recognized speaker details is displayed, else error message is displayed.
This is an entire voice recognition system with 9 functionalities
To run this program, record a few voices and store it in .wav format in the same destination as the code.