基于LPCC的病理嗓音细分识别研究.docVIP

基于LPCC的病理嗓音细分识别研究 中文摘要 语音识别技术是新世纪一门十分热门的技术，涉及多个领域，在社会生活中具有举足轻重的意义,而语音信号特征参数又是语音识别领域的重中之重，选取良好的特征参数有助于提高语音识别率 。在自动检测正常嗓音和病理嗓音时，关键是如何选出有效的特征参数、如何选择合适的识别模型使得对正常嗓音和病理嗓音进行识别得到最好的识别率。为了能实时、便利地检测正常嗓音和病理嗓音，本课题采用了线性预测倒谱系数（LPCC）声学特征参数，并利用MATLAB R2009b和Weka 3.7进行参数的提取与数据比较，在利用Weka 3.7中的7种方法分类、40种算法后，经过比较发现SimpleLogistic的相对准确率为50%; SMO的相对准确率为50%; ClassificationViaRegression的相对准确率为55.5556%。故采用ClassificationViaRegression的准确率更高，对语音识别研究更有效。 关键词：病理嗓音; LPCC参数提取; Weka 3.7 Based on LPCC pathological voice segments of research and it parameter extraction Abstract Voice recognition technology is a very popular technology in the new century, it involves many areas of great significance include social life.The characteristic parameters of the speech signal is a focus of the speech recognition, the select good characteristic parameters help to improve the speech recognition rate.Automatic detection normal voice and pathology of voice, the key is how to elect effective characteristic parameters, how to choose the appropriate recognition model makes normal voice and pathological voice recognition to get the best recognition rate.In order to be able to facilitate the detection of normal voice and pathological voice using linear predictive cepstral coefficients（LPCC） Acoustic characteristic parameters, and using MATLAB R2009b and Weka3.7 parameter extraction and comparison of data.Then it can found SimpleLogistic relative accuracy rate of 50%; for the SMO relative accuracy rate of 50%; ClassificationViaRegression relative accuracy rate of 55.5556 percent after by comparison.Therefore, the accurate rate of ClassificationViaRegression higher, its for speech recognition research more effective. Key words: pathological voice; LPCC parameters; Weka 3.7 目录 中文摘要.......................................................................................I Abstract...................................................................................................II 第一