摘要 |
PROBLEM TO BE SOLVED: To provide a method for indexing in feature vector data space for efficiently performing indexing in a vector space in which the number of dimensions, where feature vectors are not distributed uniformly, is high. SOLUTION: The method for indexing in the feature vector data space includes a step (a) adaptively constituting the approximation of the feature vectors based on the statistical distribution of feature vector data in the feature vector data space. The method performs indexing in the vector space in which the number of dimensions (dimensionality), where the feature vectors are not distributed uniformly, is high in general. Furthermore, a method like this has the advantage of easily upgrading an index when new feature vector data is added.
|