Extraction of Adventitious Sounds from Noisy Lung Sound using VMD-KLD and VMD-JSD

Lung sound (LS) provides a convenient technique to detect and discriminate the respiratory diseases. However, in realistic environments these sound recording are subject to serious noise contamination which may be addressed by Variational Mode decomposition (VMD). This paper proposes a novel adventitious lung sound extraction method that combines VMD with Kullback-Leibler divergence(KLD) or Jensen- Shannon divergence (JSD). A noisy Lung sound signal is initially decomposed using VMD into its intrinsic mode functions that have limited bandwidth. Further, their probability density function is estimated using kernel density estimator. This method aims to extract the adventitious sound from noisy lung sounds by combining only the relevant modes.