Floorplan-Context-Aware Cross-Environment Human Activity Recognition from mmWave Radar Point Clouds
Organizations: Computational Analysis and Applications Research Group, University of Ottawa, Ottawa, Canada
Abstract
Human activity recognition (HAR) models based on mmWave radar point clouds and trained in controlled open spaces often experience performance degradation when deployed in furnished rooms. We propose a Floorplan-Context-Aware (FCA) HAR enhancement method that augments the input with subject-furniture relative position information to improve cross-environment performance. Because existing mmWave point-cloud datasets lack registered floorplans, we introduce a pseudo-landmark augmentation strategy for training the FCA model. In cross-environment evaluation on the mm-ADL-Home dataset, FCA achieves 77.70% accuracy and 65.49% macro-F1, improving upon the point-cloud-only direct-transfer baseline by 10.73 and 10.05 percentage points, respectively, and exceeding the strongest evaluated adaptation baseline by 7.37 and 6.27 percentage points. At deployment, FCA requires only a one-time registration of furniture landmarks, with no target-domain activity annotations or model weight updates.
Figures & tables
| Method | Acc. | Macro-F1 | Edit | F1@10 | F1@25 | F1@50 |
|---|---|---|---|---|---|---|
| Direct transfer | 66.97 6.32 | 55.44 7.52 | 49.64 5.77 | 59.83 6.02 | 56.76 6.82 | 47.87 7.90 |
| DANN | 69.01 5.68 | 56.34 8.22 | 49.48 5.89 | 60.55 6.02 | 57.79 6.90 | 48.31 8.19 |
| Fine-tuning | 69.36 6.23 | 57.24 7.48 | 50.90 6.50 | 61.53 6.17 | 58.80 7.24 | 49.57 7.87 |
| CDAN | 70.33 7.13 | 59.22 8.31 | 51.48 6.11 | 62.74 6.11 | 59.79 7.28 | 51.36 8.20 |
| FCA-HAR | 77.70 4.05 | 65.49 4.95 | 60.70 7.38 | 71.08 5.61 | 69.59 6.02 | 61.65 7.42 |
| Method | Acc. | Prec. | Recall | F1 |
|---|---|---|---|---|
| m-Activity | 74.5 5.7 | 76.1 4.9 | 74.4 6.0 | 74.0 6.3 |
| DVCNN | 79.3 5.9 | 80.2 4.7 | 79.1 5.6 | 78.8 5.6 |
| PointNet-LSTM | 86.0 3.9 | 86.9 3.0 | 85.8 4.1 | 85.7 3.9 |
| FCA-HAR | 89.8 3.0 | 90.4 2.5 | 89.6 3.1 | 89.6 3.3 |
Human Activity Recognition
Human activity recognition classifies actions and behaviors from wearable sensors, video, radio signals, and other measurements.
See all 239 papers