METHOD AND DEVICE FOR SIMILAR DATA RETRIEVAL, AND RECORDING MEDIUM FOR PROGRAM FOR SIMILAR DATA RETRIEVAL

摘要

PROBLEM TO BE SOLVED: To improve the retrieval performance of a similar data retrieval device, which returns (k) pieces of data from high-similarity ones as retrieval results when a retrieval key is given, by suppressing an increase in distance calculation frequency in proportion to an increase in the number of pieces of data. SOLUTION: Cell spaces obtained by dividing a multidimensional space including a point that a feature variable vector indicates are hierarchically structured and feature variable vectors are uniquely manages (11, 12, 13, 14) by using bit arrays representing respective cell spaces which are made hierarchical. When retrieval is carried out, a cell space is restored (15) on the basis of the cell width at the time of the division and the distance between the point in the multidimensional space that the feature vector of the retrieval key indicates and the cell space is calculated (16); and candidate cell spaces are narrowed down (17) according to the calculated distance and a retrieval result is structured (18) as to the point in the cell space of the feature variable vector included in a candidate cell space according to the distance of the feature variable vector of the retrieval key to the point in the cell space.