经验模态分解法识别声带息肉和声带囊肿的研究

Empirical Mode Decomposition in the Research of Vocal Polyp and Vocal Cyst

为了提高声带类病理语音识别率,本文提出了一种采用经验模态分解法(Empiricial Mode Decomposition,EMD)识别声带息肉和声带囊肿的研究方法。首先采用经验模态分解法对正常语音和声带息肉类、声带囊肿类病理语音进行分解,求取语音信号的固有模态函数(IMF),经过希尔伯特-黄变换(Hibletr-Huang)变化之后,提取边际谱和特征参数用于声带类病理语音的细分。实验研究表明,采用支持向量积(SVM)边际谱和参数识别声带息肉、声带囊肿、正常语音,识别率高达90.96%。

A method based on empirical mode decomposition is proposed to improve vocal cord disease recognition rate.First,using the empirical mode decomposition method(EMD)decomposes normal voice and pathological voice of vocal cords polyp and vocal cord cyst to calculate intrinsic mode functions(IMF)of the speech signal.After Hilbert-Huang transform,we can extract the sum of the marginal spectrum feature parameters,used for segmentation of vocal class of pathological voice.The experimental results show that using support vector product(SVM)identification of the sum of the marginal spectrum parameters of three classification,including vocal cords polyp,vocal cord cyst,of normal speech recognition,the recognition rate is 90.96%.