基于声母和韵母的单字和单词识别研究

THE RECOGNITION OF SYLLABLES AND WORDS BASED ON CONSONANTS AND VOWELS

本文介绍了一个基于声母、韵母识别基础上的单字和单词识别系统。采用改进的三电平削波基音检测法。用基音周期比较法判定声调。利用共振峰参数识别韵母,用音长、短时谱参数识别声母。最后进行单字和单词的识别。本系统经一特定话者实验,单字识别率83.7％,单词识别率94.8％。

This paper presents a Chinese syllable and word recognition system based on the recognition of consonants and vowels. A modified trilevel center clipping pitch detection method is adopted in this system, and the method of three pitch comparison is used for the tone determination. After that, the System extracts the formant frequency for the vowel recognition, and the consonant length, short-time spectrum are used for the consonant recognition, The last step is to implement the recognition of syllables and words. A speaker-dependent recognition experiment has been carried out and the accuracy rate of syllables is 83.7％, the recognition score of words is 94.8％.