Aug 2026· IEEE International Conference on Embedded and Real-Time Computing Systems and Applications· pp. 218-219· 0 citations· 7 references
Abstract
Radar-based human activity recognition (HAR) is a promising solution for elderly healthcare because it enables non-contact and privacy-preserving monitoring. However, existing lightweight CNN-based methods often have limited receptive fields, while many transformer-based approaches introduce high computational overhead. To address these limitations, a lightweight Multi-Scale Inverted Residual Network (MSIRNet) is proposed for radar-based HAR. MSIRNet integrates parallel multi-scale depthwise convolutions within an inverted residual framework to efficiently capture radar activity patterns. Experimental results on a 15-subject radar dataset under leaveone-subject-out (LOSO) evaluation demonstrate that MSIRNet achieves 81.60% accuracy with a model size of only 0.97 MB and 7.29 ms inference latency on a Jetson Orin Nano.
A three-stage Adaptive CNN-Enhanced Scale-fusion Network (ACESNet) for six-channel accelerometer–gyroscope activity recognition supports a favorable performance–efficiency trade-off relative to the selected baselines under the stated protocol.
Dong-Peng Xie, Pei-Hui Yan, Yi-Fei Li et al.· Italian National Conference...· 0 citations
As Internet of Things (IoT) systems increasingly require privacy-aware sensing, radar-based human activity recognition (HAR) has become a promising option for smart homes, healthcare, and security. However, most existing methods assume a closed set of activities and often misclassify previously unseen behaviors. To add...
Kuiyu Chen, Chen Liu, Yunchao Song et al.· IEEE Internet of Things Jour...· 0 citations
The results demonstrate that combining modern spectrogram encoding, adaptive radar view fusion, and bidirectional temporal modeling provides an effective framework for continuous and distributed radar-based HAR.
The findings nonetheless illustrate the feasibility of combining principled feature selection with quantization-aware, hardware-validated deployment on embedded artificial intelligence (AI) platforms.
Mohammad Yousefi, Emine Berjin Doğan, Saeid Karamzadeh· Electronics· 0 citations
A resource-constrained CNN-GRU hybrid model for HAR on the WISDM dataset that uses convolutional layers for spatial learning and gated recurrent units (GRU) for sequence learning, enabling real-time HAR on edge devices.
H. S. Ganesha, K. G. Harsha, Roshini P et al.· International journal of com...· 0 citations
Ultra-wideband (UWB) radar is promising for human activity recognition (HAR), but its coarse micro-Doppler resolution and the scarcity of labeled data hinder practical application. To address these challenges, we propose an adaptive kinematic masking method under a simulation-to-reality (Sim2Real) framework. First, an...
Yu-Chao Guo, Nai-Ke Du, Ren-Cheng Song et al.· IEEE Signal Processing Lette...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
