腭裂语音高鼻音等级自动识别算法研究  

Automatic Hypernasal Detection Based on Acoustic Analysis in Cleft Palate Speech

作　　者：何凌[1] 袁亚南[1] 尹恒[2] 张桠童 张劲[1] 刘奇[1] 李杨[2] 

机构地区：[1]四川大学电气信息学院,四川成都610065 [2]四川大学华西口腔医院,四川成都610041

出　　处：《四川大学学报（工程科学版）》2014年第2期127-132,共6页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家自然科学基金青年基金资助项目(30900391)

摘　　要：为了对腭裂语音的高鼻音进行等级区分,提出基于声学特征参数分析的腭裂语音高鼻音等级自动识别算法,提取基于香农能量和Mel倒谱系数(Mel frequency cepstrum coefficient,MFCC)的S-MFCC作为声学特征参数,结合高斯混合模型(Gaussian mixture model,GMM)分类器实现对腭裂语音4类高鼻音等级(正常、轻度、中度和重度)的自动识别。实验结果表明,提出的自动识别算法取得了较高的高鼻音类别正确识别率,对4类高鼻音的平均识别率达到79%以上,其中,提出的S-MFCC参数取得了85%的平均正确识别率,优于传统的香农能量算法、MFCC算法,具有较高的临床应用价值。In order to detect hypemasal automatically for cleft palate patients, based on Shannon energy and Me1 frequency cepstrum co- efficient acoustic features and by combining with Gaussian mixture model classifier, an automatic hypernasal detection algorithm was proposed. The experiment results showed that the presented method achieved a good performance on the detection of four levels of hy- pemasal, such as normal, low-level, moderate-level and high-level. The average classification accuracies for four levels of hypemasal were over 79%. Moreover, the correct recognition accuracy using energy plus Mel frequency cepstrum coefficient feature set reached up to 85%. The classification of hypernasal levels has important clinical applications.

关 键 词：腭裂语音 高鼻音 香农能量 MEL倒谱系数 高斯混合模型识别器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]