论文部分内容阅读
本文提出了一种基于帧间相关特性的连续语音流的音节切分方法,采用反映相邻帧间LPC系数相关程度的帧间相关特性及其参数,进行连续语音流的分段切分,并通过时域参数对切分出的各个语音段进行音素性质标记,再根据汉语音节组成规则最后确定出音节切分及其边界。汉语数字串语音流的音节切分实验表明了该方法的有效性。
This paper proposes a method for syllable segmentation of continuous speech stream based on the inter-frame correlation feature. The inter-frame correlation feature and parameters that reflect the correlation degree of LPC coefficients between adjacent frames are used to segment the continuous speech stream. According to the Chinese syllable composition rules, the syllable segmentation and its boundary are finally determined through the time-domain parameters. The syllable segmentation experiment of Chinese digital string speech stream shows the effectiveness of this method.