Speaker adaptation algorithms based on piece-wise moving adaptive segment quantization method

A speaker adaptation method is presented based on a piece wise linear mapping of spectral code vector space into the spectral vector space of an unknown speaker. Adaptation is performed by modifying the code vectors to give a better fit to input the spectral vectors, while maintaining the local continuity of distribution of the original code vectors. Two adaptation algorithms are presented: a minimum-distortion method which provides a monotonic nonincreasing distortion for the training data and a simplified method. These algorithms are investigated in terms of distortion performance and adaptation speed. Experimental results show that the method provides much higher adaptation speed than the LBG (Linde-Buzo-Gray) algorithm and that the distortion performance approaches that of the LBG algorithm as the size of the training data increases.