lebesgue
Member level 3
I am presently working on speech recognition. I am using VQ technique. The code book is formed by using the MFCC sequences obtained by processing training speech sequence for different speakers. Suppose that during the test phase if the speaker(unknown voice utterance) signal happens to be of high pitch and assume that the unknown utterance is of one the speech of the trained speech signal, will there be a major shift in the MFCC coefficients so obtained with the high pitch values? This is important because if suppose an unknown person speaks I should be able to say that this is not an authorized person and even if the person who has authorization speaks in a low pitch or a high one I should be able to detect it with less error.please help me in this regard.:|