| 摘要 |
<p>PURPOSE: To recognize the end point with high reliability even when interference of high level is superposed on a speech signal by establishing the extreme value of a sequence of digital values derived from the speech signal. CONSTITUTION: Digital values which are previously received in order are assigned to three adjacent windows; and the 1st window (end window) contains a 1st digital number which was reached last, and the 2nd window (signal window) contains a 2nd digital number, which varies between a 1st specific value and a 2nd specific larger value. A threshold value is generated from the digital value of the 3rd window for respective values of the 2nd number, each digital value of the 2nd window is decreased by the threshold value, and the sum of the digital values is stored together with position data indicating the position of the 2nd window in the series of the digital values according to the result of a comparison with the sum of previous maximum values. Consequently, even if an extremely noisy signal is superposed on the speech signal, the start point and end point can be determined with the best reliability.</p> |
