论文部分内容阅读
本文介绍了基于连续分布型HMM的汉语连续语音声调识别方法，提出了一个适合于汉语连续语音声调识别的特征参数提取和识别方案。通过对汉语连续语音声调特点的分析，选择了8个音节单位的连续分布型HMM作为声调识别用基元模型进行识别试验，识别结果表明，10名话者1070个句子的连续语音声调识别的平均识别率是95.1％。
This paper introduces a continuous speech recognition method based on continuous distributed HMM for Chinese continuous speech, and proposes a feature parameter extraction and recognition scheme suitable for Chinese continuous speech tone recognition. Through the analysis of the characteristics of Chinese continuous speech tones, 8 consecutive syllable HMMs were selected as recognition models for tone recognition. The recognition results show that the average of 1070 continuous speech tone recognitions of 10 speakers The recognition rate is 95.1%.