农产品市场环境下语音端点检测的研究

语音端点检测直接决定了语音识别系统的精度和速度。在严重噪声情况下,对语音进行准确的端点检测是非常困难的。论文针对农产品市场环境中的语音信号,提出一种改进的子带谱熵的方法来确定语音起止点,即基于子带谱熵的方法,加入平滑机制并采用自适应阈值估计进行端点检测。本文采集了北京市农贸市场中的噪音作为环境音,进行端点检测,并与其它方法进行对比试验,即使在-15dB环境下,改进的子带谱熵的方法端点检测错误率也不高于20%,比传统方法更有效,且该方法鲁棒性较好。实验结果表明,自适应子带谱熵的方法可以很好的适应于低信噪比农贸市场环境下的语音端点检测,具有一定的应用前景。

Research on Speech Endpoint Detection Agricultural Market Environment

The speech endpoint detection directly determines the accuracy and speed of speech recognition system. In case of serious noise, speech endpoint detection accuracy is very difficult. For the agricultural market environment, an improved band-Partitioning Spectral Entropy method is proposed to determine the starting and ending point of the speech, that is, the band-Partitioning Spectral Entropy method, adding the smoothing mechanism and adopting the conservative detection end - point decision method. The paper collects Beijing city farmers market environment sound, this experiment is based on the comparison of the speech database and other methods. The experiment shows that the method is more effective and superior than the traditional method, and can detect the speech signal more accurately. Even in the low SNR, the correct rate remains stable.