A Low-complexity Noise Reduction Scheme for Target Detection in the 2D Integrated Sensing And Communication System

This paper deals with low-complexity noise reduction for linear frequency modulated continuous-wave (FMCW) signals in the multiple input multiple output integrated sensing and communication system. The conventional low-rank filter, which significantly reduces the noise, becomes impractical for two-dimensional (2D) arrays because of the huge computational burden caused by the 2D eigen value decomposition (EVD). To address the computational issue, first a low-complexity filter with two separated one-dimensional low-rank filters is proposed so that the high-complexity EVD can be avoided. Furthermore, a discrete cosine transform-based filter is proposed to achieve an efficient denoising performance without the need for accurate estimations of sample covariance matrix. The simulation results illustrate that: i) our scheme improves the denoising performance with a low complexity and it is practically feasible; ii) decreasing the rank leads to a higher output signal to noise ratio of the target, which further results in a better detection probability.