Robust Radar-Based Human Motion Recognition with L1-Norm Linear Discriminant Analysis

We present a novel radar-based indoor human motion classifier, which employs L1-norm Linear Discriminant Analysis (L1-LDA) to identify low-rank subspaces whereon Doppler signatures from distinct motions are most differentiable. In contrast to standard LDA, L1-LDA exhibits resistance against any outliers that may lie among the training data, e.g., due to mislabeling. Using real data experiments, we show that the proposed method exhibits enhanced performance compared to LDA when the training data are corrupted and similar performance under nominal training.