| 摘要 |
PROBLEM TO BE SOLVED: To reduce noise feeling and to improve quality of reproduced speech by adding an optimal index determined by a preprocessing frame to (M) preliminarily selected candidates and performing an actual selection to (M+1) preliminarily selected candidates. SOLUTION: A candidate vector j1 is outputted from an adaptive coding note 3, and k1 is obtained by performing synthesis filter processing for auditory weighting. Next, a scalar product CL of an auditory-weighted input vector n1 and k1 is calculated and (M) candidates are selected in descending order. If the scalar product CL is not smaller than the Mth one, the index and the preliminarily selected vectors are held. If searching is complete on all the candidates, the adaptive coding note in the optimal index in the former sub- frame is added to preliminarily selected (M) candidates to make (M+1) pieces of candidates. A squared-sum GL of n1 is calculated for the preliminarily selected candidate vectors and a vector making CL*CL/GL maximal is selected as the optimal adaptive coding note vector. |
