To solve the problem that the collected pulse wave signal is easily disturbed by high frequency noise and baseline drift, which make subsequent pathological research and analysis difficult and the measurement accuracy of inferior quality, an improved filtering algorithm combining variational mode decomposition and non-local mean noise reduction was proposed. Aiming at the problem that the selection of variational modal parameters has different effects on the results, the whale optimization algorithm was used to adaptively select appropriate parameters, and the modal components were screened according to the permutation entropy results. The experimental results show that the signal-to-noise ratio and root mean square error of the improved signal are better than other noise reduction methods after being processed by the filtering algorithm.