Selection of glottal excitation for low bit rate speech coder with speaker recognizability

We present a low bit rate speech coder with better speaker recognizability using the selection of glottal excitation for each speaker. In order to suitably represent the glottal excitation for each speaker, the excitation pulse determination algorithm used in multi-pulse excited linear predictive coding is adopted. In this system, 25 periodic pulses are determined for a voiced frame. A period of speaker-specific excitation pattern with only 3 pulses is chosen from the 25 pulses using a proposed pattern selection method. In the decoder, the 3-pulse pattern is smoothed using an FIR low pass filter in order to obtain a more smooth and continuous pattern. For unvoiced speech, random white noise is used as the excitation pattern. An informal listening test indicates that the speaker recognizability performance of our coder is better than that of LPC-10e.