First Heart Sound Detection for Phonocardiogram Segmentation

This paper presents an algorithm for accurate and improved detection of the first heart sound S₁ for heart sound cardiac cycle segmentation under noisy environments. The proposed algorithm integrates an S₁/S₂ selection step and an S₁ identification step. An adaptive sub-level tracking algorithm based on wavelet transform is proposed to separate the S₁ and S₂ from other components such as murmurs and noises. This is followed by a detection procedure based on Shannon energy to reject the overlapping interference so that the peaks of S₁ and S₂ can be detected. Criteria of time interval, energy and phonocardiogram (PCG) collecting position are used to identify S₁ with respect to the beginning of each cardiac cycle. Experimental results show that the proposed algorithm leads to an efficient segmentation of PCG cycle. Due to its simplicity and fast implementation, the method can be deployed clinically for further analysis and eventual use