摘要 |
PURPOSE:To suppress quality deterioration of decoded sound when the vector coding voice information which is not protected by an error correction/detection coding receives a comparatively large error. CONSTITUTION:In the decoding of a VSELP, each reliability information when a bit decision is performed for polarity information multiplying each base vector is determined. This information is successively distributed and supplied to gain generation parts 331 to 33n in the same way as the polarity information in a distribution part 36. In the gain generation parts 331 to 33n, the gain according to each input reliability information X (for instance, 1 in the case of A or more, 0 in the case of B or below, (x-A)/(A-B) between A to B) is generated. Each of these gains is multiplied by the outputs of multipliers 241 to 24n multiplying each base vector and the corresponded polarity information in multipliers 341 to 34n, the multiplication result is added and the added one is defined as a decoding vector. |