双门限算法在藏语语音音节分割中的应用分析

双门限算法是语音端点检测的一种重要方法,对藏语语音识别和处理具有重要意义。提出了用双门限端点检测技术对藏语语音进行音节分割的方法,首先根据双门限语音端点检测原理进行Matlab编程和仿真,然后结合藏语语音的音节特点和双门限算法分别在正常语速和慢语速环境下对藏语的30个辅音语音、随机抽取的双音节、三音节及句子语音进行双门限算法的音节分割和分析,实验表明双门限算法对没有太多连读音节的藏语语音和慢语速下长句的音节分割准确率较高。

Application Analysis of Tibetan Syllable Segregation with Double-threshold Algorithm

Double-threshold algorithm is the basic method for speech endpoint detection. It is essential for TibetanSspeech recognition and processing. A new method that separates Tibetan syllable with double-threshold algorithm is proposed. Firstly, based on the MATLAB platform, simulation and programming are carried out according to the principle of double-threshold speech recognition algorithm, then Tibetan syllable segmentation is tested and analyzed by combination of Tibetan syllable features with double-threshold algorithm. Tibetan speech data were tested under two speech speed environment: low speed and normal speed, divided into four groups including 30 Tibetan consonant, bisyllable, three-syllable, and random spoken sentence. The results showed that Tibetan syllable segregation has highSaccuracy with low speed long sentences and those without much connected syllables.