摘要 |
PURPOSE:To detect whether or not there is a voice in a short process time while a delay time is suppressed to <=7 msec by outputting a voiced/voiceless flag which indicates a voiced or voiceless state. CONSTITUTION:Prediction coefficients a1 and a2 are, inputted to and framed by framing units 31 and 32 at intervals of 5msec, and supplied to mean value calculators 33 and 34. The mean value calculators 33 and 34 calculate and input mean values of one frame to a voiced/voiceless state decision unit 35. The decision unit 35 sets the voice detection flag (u) to the voicelss state when the mean value of the prediction coefficients a1 and a2 is within a specific threshold range or to the voiced state when not. A hangover processor 36 performs a 100msec hangover process for the obtained result to obtain a final voice detection output (v). Consequently, the required processing time for voice detection processing is reduced to about 5msec or made short and the detector is realized with small-scale hardware. |