| 摘要 |
PROBLEM TO BE SOLVED: To correctly perform automatic level adjustment of input voice, even when high-level background noise or unvoiced sound is inputted. SOLUTION: A voice determining means 102 for determining the nature of inputted voice signals is provided with a voiced sound determining function for determining the characteristics of the voiced sound in the inputted voice signals. As a result of the determination, parts of the inputted voice which are determined as being unvoiced sound are processed, in the same way as silent signals. The voiced sound determining function calculates the autocorrelation function with respect to the pitch of a specified type from a frame and the ratio between the autocorrelation function and the frame strength and finds the pitch and the ratio, by which the ratio between the autocorrelation function and the frame strength is maximum. If the maximum value exceeds the specified threshold, it is judges as voiced sound. In this way, an appropriate gain control can be performed without needlessly increasing the gain, when high-level background noise or unvoiced sound is inputted. |
