论文部分内容阅读
该文对不同语速下,人工标注的维吾尔语连续语音语料中各音素进行共振峰频率、音长、音强的统计分析,并完成辅—元结构下的塞音、塞擦音的声学特征分析。该文通过美尔频率倒谱系数与共振峰频率等声学特征的融合及模型状态数的修改,对维吾尔语音素识别的声学模型进行了改进,并验证了不同声学特征对音素识别的影响。相比于基线系统,改进后声学模型的识别率取得一定提升。同时,利用语音学知识分析维吾尔语易混淆音素产生原因,为音素识别声学模型的进一步改进提供参考依据。
This paper analyzes the statistical analysis of the formant frequency, the pitch length and the sound intensity of each phoneme in the annotated Uyghur continuous speech corpora at different speech rates and completes the acoustic feature analysis of the consonant and fricative of the sub-element structure . In this paper, the acoustic model of Uyghur phoneme recognition is improved by the fusion of the acoustic features such as the frequency cepstral frequency and the resonance peak frequency and the modification of the state number of the model, and the effect of different acoustic features on phoneme recognition is verified. Compared with the baseline system, the recognition rate of the improved acoustic model is improved to a certain extent. At the same time, the use of phonetics knowledge of Uyghur language easy to confuse the phoneme cause, to provide a reference for the further improvement of the phoneme recognition acoustics model.