Ultra-wide band (UWB) radar is a research hotspot in the field of contactless vital sign detection. It is difficult to detect heartbeat information of human accurately at a distance more than 1 m due to the weak echo. In this paper, a variational mode decomposition (VMD) method based on time domain coherent accumulation (TDCA) is proposed to detect human vital signs such as respiration rate (RR) and heart rate (HR). Besides, radar echo of vital sign detection is modeled. The vital sign detection region (VSDR) is determined according to the standard deviation of radar echo, and a series of slow-time signals in this region are accumulated with TDCA method, so as to improve the signal-to-noise ratio (SNR) of vital signs. In order to improve the accuracy of RR and HR detection, the sliding time window combined with VMD algorithm is used to detect. Experimental results in different distances show that the proposed VMD method based on TDCA can effectively detect RR and HR. The detection distance of HR can be increased to 2 m and the accuracy of HR detection is more than 93%.