A Two-stage Classification of Heart Sounds using Tunable Quality Wavelet Transform Features

Early-stage detection of CVDs will help reduce the burden of cardiovascular diseases, which is the most significant cause of death worldwide. For the early-stage detection, the health status of the heart should be monitored frequently, which can be achieved if the user can perform monitoring of the heart without the intervention of a medical expert. To address this requirement, in this paper, a two-stage algorithm for automatic analysis of heart sounds has been proposed. In the first stage, the detection algorithm will help the user detect any abnormality at the user end. In case of any abnormality, the signal will be transferred to the medical end, and a specific disease will be identified. In the detection stage, we follow Data-preprocessing, De-noising, Segmentation, Feature Extraction and the Detection of abnormality of the signal are performed. After detection, the identification of abnormality is performed. The proposed method was applied on the dataset consisting of five categories and achieved 100% accuracy, sensitivity and specificity in the detection phase and obtained an accuracy of 98.76%, the sensitivity of 98.72%, specificity of 99.75%. The obtained results show the method’s effectiveness in detecting the abnormality and identifying a disease, which is superior to the state-of-art methods.