Analysis of Cough Sound for Pneumonia Detection Using Wavelet Transform and Statistical Parameters

Pneumonia is cause of death of millions of children under 5 years every year. Pneumonia is contagious disease which easily spread through air. In-spite of its severity the conventional methods used for diagnosis of pneumonia are costly and not easily available in resource poor regions. This paper presents a simple tool for diagnosis of pneumonia using cough sound analysis. Pneumonia patient's cough sound is collected, wavelet decomposition of the sound is performed and then statistical parameters are found out to divide the cough into pneumonia or non-pneumonia. In this paper, a crackle signal is analysed using different scales and different wavelets in CWT. For pneumonia cough samples the same steps are applied. FFT of the signal gives the frequency in required range. Then statistical parameters are calculated for original cough segments and also for CWT coefficients. Power spectral density of original signal as well as CWT coefficients is performed, in which, CWT coefficients gives required results. Threshold values of skewness and kurtosis are used to divide cough signals into pneumonia or non-pneumonia.