摘要 |
PURPOSE: A method for detecting a voice signal using temporal weights is provided to put a weight corresponding to a high frequency region on the start portion of the voice signal and a weight corresponding a low frequency region after a predetermined period of time to detect the voice signal accurately. CONSTITUTION: The method for detecting the voice signal using temporal weights includes following steps. At first, the weight corresponding to the high frequency region(A) is set(S203). The value of a timer(W-time) is set to be zero(S205). Then, high frequency high speed fourier transformation is performed(S207). The fourier transformed frequency band is processed with a filter bank process(S209). The weighted energy(EG_W) for filter bank processed frequency is calculated(S211). Then, the weighted energy(EG_W) and a predetermined value is compared(S213). Then, the voice signal is detected(S215) and the timer is incremented(S217). The incremented W_time and the predetermined time are compared(S219). A weight corresponding to the high frequency region(B) is set(S221) when W-time is greater than the predetermined time value.
|