摘要 |
PROBLEM TO BE SOLVED: To reduce the quantization error without deteriorating the quality and acoustic sense of the voice signal by obtaining the signal having strong correlation in the time base direction based on the prediction-coded signal, applying matrix quantization to it, then applying vector quantization to the signal. SOLUTION: LSP parameters for two frames are matrix-quantized by the first matrix quantization section 620, and the obtained quantization error is further matrix-quantized by the second matrix quantization section 6202 . The quantization error for two frames from the second matrix quantization section 6202 is inputted to a vector quantization section 640 constituted of the first and second vector quantization sections 6401 , 6402 . The quantization error from a matrix quantization section 620 is vector-quantized for each frame by vector quantization sections 650, 660. The obtained quantization error vector is further vector-quantized by vector quantization sections 670, 680. |