Research map / Graph learning / Knowledge and temporal graphs
Triple Spectral Fusion for Sensor-Based Human Activity Recognition
Ye Zhang, Longguang Wang, Qing Gao, Chaocan Xiang, Mohammed Bennamoun, Y Guo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add IMU fusion blocks (ablation) | TSF · Average over ten HAR datasets | macro F1 / WF1: 85.84%/92.11% -> 88.18%/93.78% | Improved |
| use adaptive complementary filter instead of traditional (ablation) | TSF · Average over ten HAR datasets | macro F1 / WF1: 84.81%/91.44% -> 88.18%/93.78% | Improved |
| add posture sensor attention (ablation) | TSF · Average over ten HAR datasets | macro F1 / WF1: 86.64%/92.52% -> 88.18%/93.78% | Improved |
| use kernel size 11 instead of 5 (ablation) | TSF IMU block · Average over ten HAR datasets | macro F1 / WF1: 87.81%/93.44% -> 88.18%/93.78% | Improved |
| use kernel size 11 instead of 21 (ablation) | TSF IMU block · Average over ten HAR datasets | macro F1 / WF1: 87.33%/93.33% -> 88.18%/93.78% | Improved |
| use kernel size 11 instead of 31 (ablation) | TSF IMU block · Average over ten HAR datasets | macro F1 / WF1: 86.88%/93.20% -> 88.18%/93.78% | Improved |
| add modality node fusion block (graph Fourier filtering) instead of MLP (ablation) | TSF · Average over ten HAR datasets | macro F1 / WF1: 86.98%/93.14% -> 88.18%/93.78% | Improved |
| add high-pass graph filters to low-pass (ablation) | TSF modality node fusion · Average over ten HAR datasets | macro F1 / WF1: 87.53%/93.25% -> 88.18%/93.78% | Improved |
| add low-pass graph filters to high-pass (ablation) | TSF modality node fusion · Average over ten HAR datasets | macro F1 / WF1: 87.62%/93.29% -> 88.18%/93.78% | Improved |
| use dynamic adjacency instead of static all-ones (ablation) | TSF modality node fusion · Average over ten HAR datasets | macro F1 / WF1: 87.32%/93.17% -> 88.18%/93.78% | Improved |
| use TSF in place of IF-ConvTransformer (add graph Fourier fusion and wavelet frequency selection) | IF-ConvTransformer · Average over ten HAR datasets | macro F1 / WF1: 86.73%/92.47% -> 88.18%/93.78% | Improved |
| use DynamicWHAR instead of DeepConvLSTM | DeepConvLSTM · Average over ten HAR datasets | macro F1 / WF1: 83.34%/90.41% -> 78.65%/86.64% | Hurt |
Closest papers by meaning
- Decomposing and Fusing Intra- and Inter-Sensor Spatio-Temporal Signal for Multi-Sensor Wearable Human Activity Recognition (AAAI 2025 · similarity 0.85)
- HARDVS: Revisiting Human Activity Recognition with Dynamic Vision Sensors (AAAI 2024 · similarity 0.83)
- Scalable Hierarchical Deep Neural Network for Time Series Analysis in Wearable Sensor-based Human Activity Recognition (AAAI 2025 · similarity 0.82)
- Generalizable Sensor-Based Activity Recognition via Categorical Concept Invariant Learning (AAAI 2025 · similarity 0.82)
- MoPFormer: Motion-Primitive Transformer for Wearable-Sensor Activity Recognition (NeurIPS 2025 · similarity 0.82)
- Limitations in Employing Natural Language Supervision for Sensor-Based Human Activity Recognition - And Ways to Overcome Them (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
