论文部分内容阅读
咳嗽是一百多种疾病的主要症状,咳嗽声分析可以为临床诊断提供极其重要的信息。利用咳嗽声信息提取咳嗽的频率和强度能够定量评估治疗效果。本文提出一种咳嗽声识别算法,首先利用小波对信号进行分解,统计咳嗽声信号和非咳嗽声信号在各个时频点上的能量分布,然后选择能量分布差异最大的部分时频点对应的能量值作为特征,最后利用线性判别分析/广义奇异值分解(Linear discriminant analysis/Generalized singular valuede-composition,LDA/GSVD)方法设计分类器。实验证明,该算法能够达到85%的识别率,且运算量较小。
Coughing is a major symptom of more than one hundred diseases and cough sound analysis can provide extremely important information for clinical diagnosis. The use of cough information to extract the frequency and intensity of cough quantitatively assess the effectiveness of treatment. In this paper, a cough sound recognition algorithm is proposed. Firstly, the signal is decomposed by wavelet, and the energy distribution of cough and non-cough sound signals at each time-frequency point is calculated. Then, the energy corresponding to the part of time- Finally, the classifier was designed by Linear Discriminant Analysis / Generalized Singular Value De-composition (LDA / GSVD) method. Experimental results show that the algorithm can achieve 85% recognition rate, and the computation is small.