Pulse wave is an important detection object for analyzing the physiological condition of human cardiovascular system. Pulse signals collected by wearable photoelectric devices often disturbed by motion artifacts due to human movement, which in turn affects the accuracy of detection results, and it is difficult to reduce the artifacts due to the fact that the main frequency components of the artifacts overlap with the frequency of the optoelectronic signals. In this paper, a pulse rate detection method combining nonlinear adaptive filtering and linear frequency-modulated Z-transform (CZT) is proposed to solve the problems of low pulse rate detection accuracy of existing pulse rate detection and insufficient resistance to motion interference. Firstly, the signal acquired by photoelectric volumetric pulse wave tracing (PPG) is preprocessed to remove high-frequency interference and baseline drift, and then the micro-electro-mechanical system (MEMS) is chosen as the reference signal to construct a nonlinear adaptive filter based on the least mean square (LMS) to remove the motion artifacts. Finally, the frequency spectrum of the signal is analyzed by CZT to calculate the pulse rate. The experiments are carried out by evaluating the performance of the experimental data collected from 10 subjects, and the results show that the method can effectively recover the original signal from the interference pulse wave signal, and the average absolute error of the measured pulse rate is 1.96 bpm, which is highly accurate and anti-interference ability.