A Wavelet-Based Denoising Under Daily Action for Respiratory Sound Extraction

Abstract

COPD is one of the most prevalent and deadliest diseases in the world today, and in 2019 it was the third leading cause of death in the world. Primary-care provider recommendations hold promise to expand the recognition of COPD in its incipient stages. For this reason, assessing the status of tracheal ventilation is crucial. To this end, for early detection and early treatment, a biological sound measurement system had been developed that measures vascular and respiratory sounds simultaneously. The system obtains the biological sound by attaching the biological sound sensor on the head. That quantitative assessment of the patient's degree of illness through changes in their tracheal sounds before and after exercise. In this study, the aim is to use wavelet analysis to isolate the noise generated during daily action for extracting the respiratory sound, including stepping conduction noise, short-time friction noise, and the measured vascular sound. Since the original signal-to-noise ratio is too small, the result evaluates the system's effectiveness by comparing the noise-free signal with the absolute noise. As a result, the amplitude and duration of noise are effectively reduced without artifacts. The signal is improved by about 30 dB on average. So, the proposed system is expected to be applied to telemedicine.