| 摘要 |
<p>This recognition system (100) is provided with a plurality of classifiers (10) and a determination unit (20). The same input data is supplied to each classifier (10), the class of said input data being unknown. Each classifier (10): uses non-overlapping feature descriptors to generate a feature vector for the input data; selects, from among representative vectors for clusters, one or more representative vectors that are the closest to the aforementioned feature vector; and outputs a classification result for the input data, said classification result consisting of the class label(s) associated with the selected representative vector(s). The determination unit (20) receives the classification results from the plurality of classifiers and applies a majority operation thereto to determine the class of the input data.</p> |
