Time domain CS kernel design for mitigation of wall reflections in urban radar

Abstract

In this paper, we use the task-specific information (TSI) metric to optimize a compressive sensing kernel for target detection behind a wall. When the target is close to the wall, strong reflections from the wall can obscure the target. Even if wall reflections are estimated and subtracted from the measurement, the dynamic range between wall and target reflections may saturate the analog-to-digital converter (ADC). Furthermore, resolving the target from the wall requires high bandwidth. In this paper, we consider the potential for using custom-designed compressive measurement kernels to mitigate these resolution and dynamic range problems. We treat wall reflections as colored noise in our Gaussian Mixture-based kernel optimization procedure, which results in custom-generated kernels that can reject the dominant wall reflections. Although this places more burden on a receivers analog components, in some scenarios it can significantly improve the situation at the ADC.