Automatic removal of lung sounds from a phonocardiogram (PCG) signal is most essential for accurately detecting and recognizing the fundamental heart sounds such as the first heart sound (S1) and second heart sound (S2). In this paper, we propose an automated lung sound removal method using the empirical wavelet transform (EWT). The proposed method consists of three major stages: the EWT based signal decomposition; the frequency based mode selection; and the signal reconstruction. The proposed method is evaluated by synthetically adding the different lung sounds available in Littmann lung sound library with the real-time recorded PCG signals from 20 volunteers. The quality of the reconstructed signals is assessed by using both objective quality assessment metrics and subjective quality test such as mean opinion score (MOS). For the performance comparison, two lung sound removal methods have been implemented based on the ensemble empirical mode decomposition (EEMD) and frequency selective filtering techniques. The objective and subjective evaluation results and the heart sound segmentation results demonstrate that the EWT based lung sound removal method outperforms the other methods. The proposed method based heart sound segmentation scheme achieves an average sensitivity (Se) of 100%, positive predictivity (Pp) of 99.22%, and overall accuracy (OA) of 99.22%. © 2019 IEEE.