Search papers, labs, and topics across Lattice.
This paper introduces a generalization-analysis framework for through-the-wall radar (TWR) human activity recognition (HAR) that addresses the challenges posed by structured distribution shifts, such as variations in person, observation view, and wall conditions. By establishing a unified source-to-target learning formulation and deriving a target-domain generalization bound, the authors provide a rigorous theoretical foundation for understanding the errors in recognition performance. Experimental results validate the theoretical insights and demonstrate the framework's practical applicability in improving HAR systems in complex environments.
Structured distribution shifts can significantly impair TWR HAR performance, but a new theoretical framework reveals how to enhance generalization across varying conditions.
Through-the-wall radar (TWR) human activity recognition (HAR) is important for non-line-of-sight indoor sensing, security monitoring, and emergency rescue. However, structured distribution shifts caused by person variation, observation-view variation, and wall-condition variation severely degrade recognition generalization, while the origin of the target-domain error still lacks a rigorous theoretical explanation. To address this issue, a generalization-analysis framework for TWR HAR is proposed in this paper. First, models for indoor human kinematics, TWR echo generation, radar image formation, feature representation, and bounded-weight neural networks are established within a unified source-to-target learning formulation. Then, the source risk, target risk, empirical risk, and admissible physical domain descriptor are defined, and a unified target-domain generalization bound is derived. Next, the structured shift term is decomposed into cross-person, cross-view, and cross-wall components, and the bound-tightening effects of physical low-dimensional representations, multi-source training, and parameter-space coverage are analyzed. Simulated and measured experiments jointly support the resulting theoretical analysis and illustrate its application value.
