Research On The Automatic Classification Of Cough

Cough is one of the most common symptoms in the clinic. It is a defensive mechanism which is caused by a sudden rapid air expulsion in the airway. The cough sound is so characteristic because of different pathogenesis that it allows classification of different types of cough. The classification of the cough could provide valuable clinical information in the assessment of patients with cough. However, the manual classification of cough monitoring recordings is a slow and tedious process, and also it may be mistaken by subjective factors. Therefore, the purpose of this paper is to research on the classification technology and try to build a special software system for the automatic classification of cough.In the paper, we classify the cough as the wet cough and the dry cough. This category has been showed very efficient in the automatic classification.The paper firstly gives a summary of the research on the speech recognition and cough analysis at home and abroad. Based on the characteristic of cough, we proposed a cough classification system using the MFCC and DTW. After studying the differences between speech and cough, we found the LPCC which is based on the vocal tract models is not appreciating for the cough classification. So we use the MFCC as the character parameters. In order to improve the dynamic performance, we added the first difference to the standard MFCC. Meanwhile, because of the possibility of degradation due to the inconsistent environment between the training and classifying, we proposed an improved algorithm to suppressing the noise. The result shows that it can greatly enhance the robustness of the system. In the template matching part, we use the DTW to accommodate the differences in the duration between different cough, and we also introduced a loose ratio in the start-end point restriction to reduce the sensibility with the endpoint detection. In the DTW algorithm, we use the modified K-Means algorithm to train the template to make sure the stability of the template.The paper has finally finished the research on the automatic classification of cough, and has designed a software system based on the proper technology. We used the system to perform a lot of experiment and the result showed a high accurateness in the classification between wet and dry cough.