Respiratory sounds denoising using wavelet packets

Lung sound analysis is made difficult by the inevitable noise contamination. Conventional frequency filtering is usually ineffective, therefore the use of other methods like those based on the wavelet transform is needed. In this paper, we propose a new technique for denoising lung sound This technique is based on Donoho's method and consists in modulating the wavelet packet coefficients before the thresholding operation and demodulating them after. By using this method, we can reduce the noise without distortion of the original sound. The results obtained with normal and pathological sounds show a very high efficiency.