In this paper, a soft threshold method based on a wavelet transform, and a decomposition reconstruction method, are proposed to preprocess photoplethysmograph (PPG) signals. First, the soft threshold method is applied to remove the high-frequency noise. The wavelet decomposition and reconstruction method are then used to adjust the baseline drift in the PPG signals. The aim is to obtain a higher signal-to-noise ratio and smoother PPG signals. Finally, the validity of the proposed signal preprocessing method is verified using the general indicators of RMSE, the smoothness, and the starting point variance.