PURPOSE: To decrease the quantization error by constituting each gain multiplier of the number of quantization points in which a noise characteristic derived in advance by an amplitude distribution in each shape vector direction and the gain of quantization.
CONSTITUTION: A code book 1 is constituted of shape vectors 21, 22, and to vectors (u), (v) obtained by executing auditory sense weighting through transducers 31, 32 to each vector 21, 22, respectively, each quantization point number and the gain of a quantization value are given by gain multipliers 41, 42. That is, two gain multipliers 41, 42 are constituted of gain multipliers 411-416 of six gains +0.7, +0.5, +0.2, -0.2, -0.5 and -0.7, and gain multipliers 421, 422 of two gains +0.6, -0.6, respectively. In the cane the gain multipliers are constituted in such a way, each gain multiplier 411-416 and 421, 422 gives an efficient quantization point.
MATSUO NAOJI
ISEDA HIDEHIRA
AMANO FUMIO
