| 摘要 |
In a speaker distance calculation part of a speaker recognition apparatus, a quantization distance is obtained between a voice feature vector of a voice feature vector sequence produced from a voice of a speaker to be recognized and a representative vector of a codebook. The voice feature vector is quantized based on the quantization distance. And a quantization distortion is obtained by using a high order voice feature vector group of the voice feature vector sequence. In a recognition part of the speaker recognition apparatus, speaker recognition is executed based on the quantization distortion, which is for example, an average value of a plurality of the quantization distortions.
|
