Automátic Speaker Identification by means of mel cepstrum, wavelet and wavelet packet

H. M. TORRES; H. L. RUFINER

Chicago, EEUU.

Conferencia; World Congress on Medical Physics and Biomedical Engineering; 2000

IEEE Engineering in Medicine and Biology Society

The present work consists on the use of Delta Cepstra Coeficients in Me1 scale, Wavelet and Wavelet Packets Transforms to feed a system for automatic speaker identification based on neural networks. Different alternatives are tested for the classifier based on neural nets, being achieved very good  performance for closed groups of speakers in a text independent form. When a single neural net is used for all the speakers, the results decay abruptly when increasing the number of speakers to identify. This takes to implement, a system where there is one neural net for each speaker, which provided excellent results, compared with the opposing ones in the bibliography using other methods. This classifier structure possesses other advantages, for example, add a new speaker to the system only requires to train a net for the speaker in question, in contrast with a system where the classifier is formed by a single great net, which should be in general trained completely again.