Classification of Normal/Abnormal Heart Sound Recordings Using Time –Frequency PCG Signal Analysis

The heart’s acoustic signal produced by its mechanical activity can provide useful information on the condition of heart valves. The heart sound auscultation, i.e. listening to the heart sounds with a stethoscope, is therefore a primary method for evaluating the cardiovascular function. This method has advantages of being fast, inexpensive, easy to use and noninvasive. On the other hand, due to the transient and non-stationary nature of PCG signals and auscultatory limitations, the correct medical diagnosis based on the heart sound through a stethoscope requires a lot of expertise and needs referral of the patient to a cardiologist.  This is not only time-consuming but also imposes a financial burden on the medical system. Thus, automated detection and analysis of the recorded heart sound auscultation has received a lot of attentions in recent years. Even, this was put to the challenge by the PhysioNet/CinC in 2016.This research is based on the non-stationary nature of PCG signals and proposes a new method based on time–frequency analysis of such signals with the aim to classify heart sounds into normal and abnormal sounds. The proposed methodology uses time-frequency features and two classifiers; AdaBoost and CNN. The publicly available 2016 PhysioNet/CinC 2016 Challenge database was used to evaluate the performance of the proposed method using a leave-one-out cross validation. The experimental results show that the proposed method performs very well and has 1.5% higher sensitivity compared to the best existing method.