Real-Time Point-Cloud Detection for Human Activity Sensing Based on a 77-GHz Stepped MIMO Radar

In this paper, we propose a real-time point-cloud detection method using adaptive approximate Kaiser filter (AAKF) weighted 4D fast Fourier transform (FFT) based on a 77-GHz frequency modulated continuous wave (FMCW) radar. This method can be used for the human sensing, achieving the spatial position and general posture detection accurately and quickly. This method consists of two main parts. First, combining the adaptive approximation strategy and table-lookup method, an AAKF algorithm is proposed to suppress the sidelobes flexibly and quickly. The second part involves a multiple input multiple output (MIMO) multi-channel signal sparse separation method, enhancing the angle estimation accuracy and robustness. It reconstructs the weight matrix with lower time consumption and lower resource usage through mathematical relational mapping, overcoming the inherent dependence of beamforming technology on uniform arrays. To verify the proposed method, a customized stepped-arranged 12T16R MIMO radar has been designed. The experimental results show that the AAKF algorithm can reduce the computational time by nearly 10 times compared to the traditional Kaiser filter algorithm and increase the signal-to-interference ratio (SIR) by 12.7 dB, while the separation method cuts the time from 3.99 ms to 1.51 ms, just 37.8% of the existing method. This point-cloud detection method can achieve excellent detection effect, and have breakthroughs in both the detection accuracy and computational efficiency, providing a more robust technical solution for the high real-time engineering applications hopefully.