recognition device and recognition method and device and learning method, the study notes

摘要

the author of the voice data and the case, and will, in the form of image data is obtained, and the noise data and plural pieces of data in the form of the output from the input part (10). the data processing part (20 to 23) and the analysis on characteristic parameter is obtained. the classification of the data structure (24) of the data is obtained from the parameters of the classification and the classification of (25). the parameters of the processing part (26), (20 to 23), the output of the feature parameter from the parameter to form integration. adaptive decision part (27), the classification part (25) of the classification results is selected in response to the table.the speaker's voice is generated in the selected table and the stored reference parameter, and the parameter group (26), the output of the integration according to the parameters. the voice of the recognition rate is improved.