摘要 |
PROBLEM TO BE SOLVED: To enable extraction of only those frequency components with respect to a feature sound. SOLUTION: A wavelet transforming means 13 separates the low-frequency components from high-frequency components, by making the signal corresponding to the sound wave of the object acquired by a voice input means 11 as the first input signal, and carrying out wavelet transformation to the input signal. If the feature sound, fitting to the purpose, is contained in the low-frequency components separated by the wavelet transforming means 13, a monitoring means 14 makes the low-frequency components as the next input signal of the wavelet transformation means 13, and extracts the frequency band of the input signal of the wavelet transforming means 13, when the feature sound no longer is contained. The input signal to the wavelet transforming means 13, when the feature sound no longer is contained is inputted into a filtering means 15, in which the frequency band extracted by the monitoring means 14, is set as a passage frequency band. For the signal which passed the filter means 15, a frequency spectrum is detected by a vector-extracting means 16. COPYRIGHT: (C)2005,JPO&NCIPI
|