| 摘要 |
Speech signal presence is decided based on the input signal exceeding either of two thresholds: one, a fixed threshold (TF) set at an arbitrary level relatively high above anticipated noise; the other, an adaptive threshold (TL) which idles slightly above noise. If the input signal rises above the idling threshold of TL, speech presence is indicated. If the input signal continues to rise (i.e., amplitude-time slope positive), the presence indication continues. If the input signal level falls, the adaptive threshold is adjusted (TL=BT+D, where for example B=1, D=5 and T=the current signal sample average value). Hangover is controlled by the amount of time the input signal exceeds the threshold TL. Speech presence is also indicated by the input signal exceeds a third threshold (TH) which is also adaptive, and idles at a relatively high level above noise.
|
