Abstract: Addressing the issues that static clutter and respiratory harmonic interference noise exist in non-contact vital signs detection of millimeter wave radar, an improved coatis optimization algorithm based variational mode decomposition (ICOA-VMD) noise suppression algorithm is presented in this paper. The coatis optimization algorithm uses chaotic population initialization and adaptive function distribution to improve the population diversity and global search ability of the algorithm. So, ICOA is used to search the optimum fitness parameters of VMD and determine the penalty parameters and the number of components in this paper. Then, the heartbeat signal is reconstructed to remove the interference and noise of heartbeat signal. The experimental results show that the ICOA-VMD method has the characteristics of fast convergence and high precision, and evaluation and time domain analysis through signal-to-noise ratio and mean squared error show that the proposed algorithm has better performances than wavelet transform and empirical mode decomposition. The average accuracy of the proposed method can reach 95.40% for the heart rate detection of different subjects in the conventional environment at different distances.