Millimeter-wave radar noncontact vital sign sensors are widely used in home care and clinical monitoring. However, in practical applications, these sensors face multi-source dynamic interferences such as motion artifacts, environmental vibrations, and electromagnetic noise, leading to a severe degradation in estimation accuracy. To address this issue, we propose a Chameleon Swarm Algorithm-assisted Optimized Successive Variational Mode Decomposition (CSA-SVMD) algorithm in this paper. The algorithm constructs a fitness function based on the time-frequency domain characteristics of vital sign signals, utilizes the Chameleon Swarm Algorithm to achieve the adaptive optimization of the balancing parameter in Successive Variational Mode Decomposition, and finally realizes the accurate estimation of vital signs by combining frequency priors and magnitude-squared coherence. This approach ensures estimation accuracy while significantly enhancing robustness against multi-source dynamic interference. Experimental results based on the BMR2 vital sign monitoring board show that the CSA-SVMD algorithm achieves a root mean square error (RMSE) of about 2 beat per minute (bpm) for respiratory rate and about 3 bpm for heart rate at a distance of 0.5 m across all tested subjects. Although RMSE increases with measurement distance, the algorithm maintains a stable performance degradation trend in multi-source dynamic interference scenarios.