基于区别特征系统的连续语音识别模型研究

The Research of Continuous Speech Recognition Model Based on Distinctive Feature System

黏着性语音的元辅音区别性特征对口语环境下的连续语音识别影响很大.通过维吾尔语和谐语音现象下的发音规律及其规则,采用语言学、实验语音学技术和数字语音信号处理等方法,研究和谐发音现象下的元音、辅音区别特征系统,建立扩展元辅音集,并以此为基础,开展基于扩展元辅音集的连续语音识别研究,进一步通过其声学模型比较研究,构建连续语音识别模型,为维吾尔语口语环境下的连续语音识别提供一种新方法.

The distinguishing feature of vowels and consonants of adhesion speech has great influence on the Continuous speech recognition in oral environment. We explore the pronunciation rules under Uighur harmonious voice phenomenon. We can adopt the linguistics,experimental phonetics and digital speech signal processing technologies to make focused research on the distinctive feature system of vowels and consonants under it harmonious pronunciation phenomenon. Then, we can construct the Uighur expanded vowel-consonant dataset and carry out the research on the continuous speech recognition based on the expanded vowel-consonant dataset. Further, we construct the continuous speech recognition model by performing the comparison of acoustic models and provide a new method for the Uighur continuous speech recognition in oral environment.