This paper aims at the speaker‐independent very‐low bit rate speed coding and proposes the speaker adaptation algorithm based on the piecewise movement in the speech spectrum space. In the proposed method, the universal codebook, which is constructed beforehand based on the features of the spectral distribution of an unknown speaker, is adapted to the speaker. The adaptation is based on the process which moves the center of gravity of the universal codebook for several clusters. Two algorithms are proposed for the adaptation. One is a simplified method and the other is a minimum distortion method. The minimum distortion method ensures theoretically that the quantization distortion does not increase for the speech data for adaptive training. A coding experiment was made for the spectral segments of the actual speech. It was shown that the proposed adaptation method has a higher adaptation speed compared with the traditional LBG method. It was shown also that the minimum distortion method converges to the same distortion as the LBG method.
ASJC Scopus subject areas
- Theoretical Computer Science
- Information Systems
- Hardware and Architecture
- Computational Theory and Mathematics