摘要 |
PROBLEM TO BE SOLVED: To detect a speech section as the speech section of a whole channel, when the speech section exists in at least one channel in input signals of n-channels. SOLUTION: In a speech section detection device, each input signal of n-channels is framed and stored in a memory (S101). Regarding a signal sample stored in the memory for each channel, a result (VAD flag) in which the signal sample is in a speech section or in a non-speech section, and time (VAD flag determination time) when the VAD flag is determined, are output. The VAD flag determination time (head time) of the earliest time in each VAD flag determination time, is searched for, and for each VAD flag, when there is at least one for indicating the speech section, it is determined that an integrated detection result is the speech section, and when all indicate the non-speech section, it is determined that the integrated detection result is the non-speech section, and the head time and the integrated detection result is output. COPYRIGHT: (C)2009,JPO&INPIT
|