基于能零比与峰谷点的汉语语音音节分割算法

Chinese speech syllable segmentation based on energy-zero ratio and peak-valley point

为了提高连续语音的音节分割效果以及在噪声环境中不降噪情况下音节分割的效果,提出了一种基于能零比与峰谷点的汉语语音音节分割算法。该算法的主要思想是融合两种算法确定的两轮音节分割边界,第一轮基于能零比确定有话段与无话段的分割边界,第二轮基于时域波形包络的波峰波谷确定该段语音的音节分割边界。实验结果表明,该算法在无噪环境中的语音音节分割效果优于传统方法;在噪声环境中不降噪的情况下仍能保持较高的准确率。

In order to improve the syllable segmentation effect of continuous speech and the effect of syllable segmentation without noise reduction in noisy environments,this paper proposes a Chinese speech syllable segmentation algorithm based on energy-zero ratio and peak-valley point.The main idea of the algorithm is to combine the two-round syllable segmentation boundary determined by the two algorithms.The first round determines the segmentation boundary between the voice and no voice based on the energy-zero ratio,and the second round determines the segmentation boundary based on the peak and valley of the time domain waveform envelope.Experimental results reveal that the effect of speech syllable segmentation in the noiseless environment is better than the traditional method and generate good robustness.In noisy environments,the algorithm can still maintain high accuracy without noise reduction.