LiDAR-based Human Activity Recognition through Laplacian Spectral Analysis

2025-09-27

AI Summary

To address privacy concerns, illumination sensitivity, and limited interpretability in LiDAR point cloud-based human activity recognition (HAR), this paper proposes a lightweight geometric feature method grounded in spectral graph theory. Point clouds are modeled as ε-neighborhood graphs, and spectral features—namely, eigenvalue distributions and statistical descriptors of dominant eigenvectors of the normalized Laplacian matrix—are extracted as pose-invariant descriptors. Temporal dynamics are captured via sliding-window aggregation, followed by classification using SVM or random forests. The approach avoids end-to-end deep learning and explicit skeleton estimation, yielding compact, physically meaningful features. Evaluated on the MM-Fi dataset, it achieves 90.3% accuracy across 27 daily activities performed by 40 subjects, and 94.4% on 13 rehabilitation tasks—significantly outperforming skeleton-based baselines. The method delivers high accuracy, strong robustness to environmental variations (e.g., lighting), and enhanced model interpretability.

Technology Category

• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Computer Vision: Video Understanding & Activity Analysis
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Security and Privacy: Large-scale security measurements

Abstract

Human Activity Recognition supports applications in healthcare, manufacturing, and human-machine interaction. LiDAR point clouds offer a privacy-preserving alternative to cameras and are robust to illumination. We propose a HAR method based on graph spectral analysis. Each LiDAR frame is mapped to a proximity graph (epsilon-graph) and the Laplacian spectrum is computed. Eigenvalues and statistics of eigenvectors form pose descriptors, and temporal statistics over sliding windows yield fixed vectors for classification with support vector machines and random forests. On the MM-Fi dataset with 40 subjects and 27 activities, under a strict subject-independent protocol, the method reaches 94.4% accuracy on a 13-class rehabilitation set and 90.3% on all 27 activities. It also surpasses the skeleton-based baselines reported for MM-Fi. The contribution is a compact and interpretable feature set derived directly from point cloud geometry that provides an accurate and efficient alternative to end-to-end deep learning.

Problem

• Developing privacy-preserving human activity recognition using LiDAR
• Creating graph spectral features from point clouds for classification
• Providing interpretable alternative to deep learning for activity analysis

Innovation

• Graph spectral analysis for LiDAR point clouds
• Laplacian eigenvalues and eigenvectors as pose descriptors
• Compact interpretable features from point cloud geometry