With the aging society and the increasing demand for health management, low-power millimeter-wave radar health monitoring systems have become a research hotspot due to their advantages such as non-contact and anti-interference. However, traditional real-time signal processing algorithms have high computational complexity and large resource consumption, which makes it difficult to meet the operating requirements of low-power devices. To this end, this paper proposes a lightweight algorithm design scheme for real-time signal processing based on a low-power millimeter-wave radar health monitoring system. First, the millimeter-wave radar echo signal is feature extracted, and the improved sparse reconstruction algorithm is used to reduce the amount of data processing; then, an adaptive filtering algorithm is used to optimize the signal quality and reduce noise interference; finally, a hardware-friendly parallel computing architecture is designed to improve the execution efficiency of the algorithm. Experimental results show that the lightweight algorithm reduces the amount of FFT calculation by 60% through SSP, and the matrix operation complexity decreases by 75% after IPD replaces MUSIC. Finally, the optimization effect of reducing the amount of calculation by 40% (12.8 → 7.7 GFLOPs), shortening the processing time by 35% (82 → 53 ms), and reducing the memory usage by 32% is achieved. In the multi-object parallel test, the total power consumption of the system only increases by 0.9 W, which verifies the scalability advantage of the hardware pipeline design.