论文部分内容阅读
咳嗽识别在临床上具有重要的诊断指导意义。针对咳嗽频谱能量的分布特点,本文提出了一种新的梅尔(Mel)频率倒谱系数(MFCC)提取方法。将咳嗽频谱划分为若干个频段,采用主元分析方法计算各频段的能量强度系数,根据强度系数的插值曲线分配滤波器个数,设计Mel刻度上非均匀分布的滤波器组进行MFCC特征提取。基于隐马尔可夫模型(HMM)的咳嗽识别实验表明,该方法可以有效改善咳嗽识别的效果。
Cough identification clinically important diagnostic guidance. In view of the distribution of cough spectrum energy, this paper presents a new Mel Frequency Cepstral Coefficients (MFCC) extraction method. The cough spectrum is divided into several frequency bands. The principal component analysis (PCA) method is used to calculate the energy intensity coefficients of each frequency band. The number of filters is allocated according to the interpolation curve of the intensity coefficient. A MFCC feature extraction is performed on the filter banks with non-uniform distribution on the Mel scale. Cough recognition experiments based on Hidden Markov Model (HMM) show that this method can effectively improve the effect of cough recognition.