摘要 |
<P>PROBLEM TO BE SOLVED: To provide a retrieval device etc. for retrieving a feature vector similar to a query feature vector from a set of high dimensional feature vectors, with a small amount of codebook information. <P>SOLUTION: A learning unit in a retrieval device includes residual codebook generation means that generates an n number of residual clusters by deriving residual vectors f<SP POS="POST">r</SP>between representative vectors f<SB POS="POST">n</SB>and learned vectors f (S71), generates sets of residual partial vectors Rn,t by dividing the vectors f<SP POS="POST">r</SP>into T pieces (S72), assigns label values m in a residual codebook identifier table T[n][t] (S73), generates an m-th residual codebook using all sets of residual partial vectors Rn,t to which the values m are assigned, for each value m (S74), and updates the table T[n][t] using an m'-th residual codebook which results in the smallest quantization error when quantizing using each of the m number of residual codebooks (S75). <P>COPYRIGHT: (C)2013,JPO&INPIT |