| 摘要 |
PURPOSE:To shorten a quantization processing time by calculating the distortion of an input vector from an original representative vector with which a selected candidate vector is generated, and limiting representative vectors to be searched for and performing quantization. CONSTITUTION:Auxiliary vectors U1, U2, U3... are generated previously for all representative vector Y so that only large amplitude components of the representative vectors Y1, Y2, Y3... are left. Then plural auxiliary vectors U which have small distortion between the input vector X and auxiliary vectors U1, U2, U3... are selected. Then the distortion of the input vector X from only the original representative vector Y with which the selected candidate auxiliary vectors U are generated is calculated and only the representative vectors Y to be searched for are quantized. Thus, the quantization processing time can be shortened. |
