摘要 |
PROBLEM TO BE SOLVED: To learn sound recognition object data at every sound dial calling, to reduce the effects due to a peripheral environment condition and to further improve sound recognition property. SOLUTION: A sound telephone book/memory part 19 accumulates sound data which is sound-inputted. A similarity discriminating part 21 compares the content of input sound data which is newly sound-inputted with sound data accumulated in the sound telephone book/memory part 19, and extracts similar sound data having high similarity with the content of input sound data among accumulated sound data. A standard pattern generation part 22 learns and updates accumulated sound data, when the contents of input sound data having high similarity and similar sound data are matched.
|