Real Time Lung Sound Separation from Cardiac Sounds by Adaptive Algorithm Technique

K. Sathesh, S. Rajasekaran, P. Jayapal Reddy, T. Manasa, G. Mythresh

Abstract: For reliable voice acquisition, active noise control (ANC) can be a good option. Additionally, this method has the potential to reduce the Lombard effect. The well-known Filtered-x Least Mean Squares adaptive method is the most extensively employed in broadband active noise cancellation (FxLMS). We investigate an alternative to the FxLMS algorithm that aims to solve its occasionally delayed convergence without sacrificing cancelling capability. The ALE-FxLMS system is an option provided here, in which an Adaptive Line Enhancer (ALE) is utilised as a decorrelating step for the FxLMS algorithm. The single-channel example (one reference signal, one actuator, and one error sensor) is presented and analysed, as well as three potential system extensions to the multiple channel situation. Without decorrelating pre-processing, the suggested system has been shown to provide faster convergence with reference to a single FxLMS. Since no frequency component is weighted higher than the others when using a white reference signal, the FxLMS method is projected to be faster to convergence. The ALE-FxLMS system is described in this paper, in which an Adaptive Line Enhancer (ALE) is employed as a decorrelating pre-processing stage for the FxLMS algorithm. The ALE-FxLMS system intends to increase computational complexity while improving the convergence of the entire adaptive system. The single-channel scenario is investigated, as well as the system's expansion to the multiple-channel case, where strongly correlated reference signals can be established. Three possible generalisations of multiple channels are shown. The performance of these systems is evaluated without pre-processing, using the single FxLMS as a reference.

Keywords: Heart sound signal, HSS, Lung sound signal, LSS, Adaptive line enhancer, ALE, Filtered Least mean square, FxLMS, Active noise control, ANC