Extracting feature parameters
no vote
%In VQ based speaker recognition system, vector quantization plays a dual role. In the training phase, the feature parameters extracted by each speaker are classified to generate codebooks composed of different codewords. In the recognition (matching) stage, we use VQ method to calculate the average distortion measure (the system uses Euclidean distance measure when calculating the distance d), so as to judge who the speaker is. This code is to train the audio, that is to extract feature parameters