Noisy voice endpoint detection method based on multi-resolution KL divergence and voting mechanism

The invention discloses a noisy voice endpoint detection method based on multi-resolution KL divergence and a voting mechanism. The noisy voice endpoint detection method comprises the following steps: framing sampled voice signal sampling points according to a time sequence; carrying out frequency band division on the current frame according to different granularities, and obtaining a plurality of sub-bands under each granularity; respectively calculating the KL divergence of the sub-band energy distribution under different granularities, and comparing the KL divergence of each granularity with a first threshold value to obtain a discriminant value of each granularity; and obtaining a comprehensive value according to the judgment values of different granularities, and comparing the comprehensive value with a second threshold value to obtain a voice and noise judgment result. The problem that noise is misjudged as voice due to fluctuation of a noise spectrum in adjacent sub-bands is effectively solved, so that