| 摘要 |
<p>PURPOSE: To improve the physical S/N of an input signal and to improve the clarity of the signal considering a human sensation by using a frequency analysis method the more lowered in frequency resolution for higher frequency band. CONSTITUTION: An input voice is inputted to a frequency analysis part 1, and frequency analysis are performed at both of the time when a voice is included and the time when the voice rests (when only a noise). As the frequency analysis method, e.g. a wavelet conversion is adopted. A voice emphasis is performed based on the analysis result obtained by the frequency analysis part 1. Since the frequency resolution is more lowered in the high frequency band so as to match with the human sensation by the wavelet conversion method and a geometric frequency analysis method, etc., and on the other hand, the analysis method more improved in space resolution or time resolution in the high frequency band is used, not only the physical S/N, but also the clarity in the human sensation such as monosyllabic clarity in the case of the voice emphasis is improved.</p> |
