Abstract: A heart sound extracted and processing algorithm was studied for smart clothing based on fiber grating sensors. Combining Hilbert-Huang transform (HHT) with wavelet threshold denoising, the heart sound denoising algorithm was proposed to extract the useful heart sound signal from wavelength demodulated signal. The peak points and start and end points of heart sound were obtained respectively from the envelopes extracted by mathematical morphology method based on the line and cosine structuring elements, and then the characteristic parameters of the heart sound were calculated. Using those parameters, normal or abnormal heart sounds can be identified. The experimental results show that the algorithm can effectively remove the breathing interference and noise mixed in the wavelength demodulated signal. In the tests of 20 cases of normal heart sounds and 8 kinds common abnormal heart sounds, all of the recognition results are correct. The algorithm is characterized by easy realization and high recognition rate, and it is significant to the development of the fiber optic sensing smart clothing and the early diagnosis of heart disease.