Abnormal heart sound signal recognition based on fiber-optic EFPI sensor

We propose to use Extrinsic Fabry–Perot Interferometric (EFPI) acoustic sensors to detect heart sound signals in order to eliminate the adverse effects of electromagnetic interference. A fiber-optic EFPI acoustic sensor based on graphene membrane is produced. Its mechanical sensitivity is 760 nm/Pa at 630 Hz. By analyzing the time–frequency domain, we have demonstrated the possibility and effectiveness of using fiber-optic EFPI sensors for heart sound measurement. Mel-Frequency Cepstral Coefficients (MFCCs) extracted from heart sound signals are subsequently put into two deep learning architectures: CNN+GRU and CNN+LSTM , which are utilized for the identification of abnormal heart sounds. Remarkably, these neural networks achieve recognition accuracies of up to 98.65% and 99.51% respectively, highlighting their robust performance in this application. • The fiber-optic EFPI acoustic sensor is used to detect heart sound signals to avoid EMI. • MFCC feature extraction is used to analyze nonlinear frequency components of heart sound signals. • CNN+LSTM and CNN+GRU networks are applied to recognize abnormal heart sound signals.