Abstract: This paper proposed a signal process and feature extract algorithm of human’s pulse wave signal collected by fiber Bragg grating (FBG) sensor. This paper compares the pulse wave signal collected by FBG sensor with the photoplethysmography (PPG) pulse wave and then figures out the characteristics of FBG pulse wave. A wavelet threshold de-noising combined with modified mathematical morphology is proposed, which is used in the signal de-noising works, and the length of mathematical morphology structural element is automatically selected based on the pulse cycle that can improve the effect of baseline de-noise. The feature extraction method proposed in this paper can improve the measuring accuracy of peak and start points of the pulse wave. According to the result of experiment, the Signal-to-noise Ratio(SNR) of the output pulse wave was twice as much as the SNR of the input pulse wave, and the accuracy of extracted peak and start points were over 97.2% and 97.6%. With simple structure and ease of implementation, this method is significance for the research of FBG smart clothing and the detection of pulse characteristics.