LITEWAY: LIghtweight HAR via Temporal Efficient highWAY

Edgeefficient human activity recognition using a quantized patchbased transformer.

Human Activity Recognition (HAR) systems using deep learning have shown significant promise; however, deploying such models on edge computing systems remains challenging due to constraints in inference latency, memory footprints and computational capacity. This study proposes a lightweight, end-to-end patch-based Transformer architecture designed for efficient HAR for resource constrained edge environments, evaluated through an architectural ablation study and multiple quantization strategies using TensorFlow Lite. We conducted extensive experiments by varying patch lengths and Transformer encoder depths in order to evaluate the impact on classification accuracy, latency and model complexity. The results demonstrate that moderate temporal patching achieves an effective balance between temporal representation learning and computational efficiency. Experiments have been conducted on two widely open-source benchmark datasets, WISDM and PAMAP2, where raw sensor signals are pre-processed through segmentation, normalization, and overlapping windowing before being fed into a compact Transformer model with patch embedding and multi-head self-attention for feature extraction. The trained models are converted into FP32, FP16, INT8 dynamic range, and INT8 full formats to evaluate trade-offs between different metrics such as accuracy, model size, and inference latency. The baseline model achieves 96.77% accuracy on WISDM and 98.48% on PAMAP2, with consistently high macro F1-scores. Among all quantization variants, TFLite-INT8-Dynamic reduced model size by 95.55% and 89.80% for WISDM and PAMAP2, respectively and drastically improved inference latency by around 99% for both datasets with minimal accuracy degradation. These findings demonstrate that the proposed patch-based Transformer model for HAR has achieved an effective balance between recognition performance and computational efficiency, which indicates its strong deployment feasibility for resource constrained edge environments through TensorFlow Lite benchmarking and can provide a scalable solution for edge intelligence applications.

CoDAT: Collaborative Dual-Attention Transformer with Low-Cost Temporal Modeling for Efficient Edge Action Recognition

Real-time human action recognition on Internet-of-Things (IoT) edge devices requires models that capture rich spatio-temporal cues within strict latency, memory, and power envelopes. Current 3D CNNs, video transformers, and shift-based ViT deliver high accuracy but come at computational costs that preclude edge IoT deployment. This paper proposes CoDAT, a Collaborative Dual-Attention Transformer that replaces conventional multi-head attention with a lightweight dual-branch module: Spatial Convolutional Attention (SCA) for local aggregation and Strided Single-Head Attention (SSHA) for global context. SSHA jointly compresses the spatial resolution and channel dimensions of the query, key, and value tensors via stride-based sparse projection, then fuses the resulting global and local features at a markedly reduced cost. To enable temporal communication across frames, a parameter-free TShift module is embedded in each block. Extensive experiments on Jetson AGX Orin and Raspberry Pi 5 demonstrate that CoDAT achieves an energy-accuracy balance in both image and action recognition. On ImageNet-1K, CoDAT-M runs 2x faster than EfficientViT384 and FastViT-S12 at comparable accuracy, and CoDAT-L matches ViT-S with 3x fewer parameters at 2x higher throughput. On Kinetics-400 and MA-52, CoDAT achieves competitive Top-1 accuracy against state-of-the-art CNN, transformer, and hybrid baselines while running up to 2.9x faster than VSwin-T, 2x faster than ViT-Temporal-Shift variants, and 5x faster than UniFormer-B. On UCF-101, CoDAT-S384 matches TokShift and LAPS while being 6x faster and requiring up to 13x fewer FLOPs, establishing an efficiency-accuracy balance for real-time action recognition in edge IoT perception systems. Code is available at https://github.com/novendrastywn/CoDAT .

A Lightweight CNN-GRU Model for Human Activity Recognition with Efficient Edge Deployment Using TFLite

Wearable sensor-based human activity recognition (HAR) has become increasingly popular for applications in health monitoring, fitness, and smart living. But the use of deep learning models on edge devices is still challenging due to limited memory and computational power. In this paper, we develop a resource-constrained CNN-GRU hybrid model for HAR on the WISDM dataset. This architecture uses convolutional layers for spatial learning and gated recurrent units (GRU) for sequence learning. For deployment on the edge, the model is quantized to TensorFlow Lite (TFLite) using float16. Our experiments show that the model achieves an accuracy of 94.07%, while the size of the model is substantially smaller and suitable for deployment on edge devices. Importantly, the TFLite model maintains the same accuracy as the original model, ensuring its suitability for real-time deployment. The extensive assessment through confusion matrices, ROC curves and classification metrics confirms the effectiveness of the model across various activities. The proposed approach offers a balance between accuracy and computational efficiency, enabling real-time HAR on edge devices.

CGAC: A Convolutional Bidirectional GRU Network with Temporal Attention for WiFi CSI-Based Human Activity Recognition

WiFi channel state information (CSI) can characterize wireless-channel variations induced by human activities without directly capturing identifiable visual content, providing a contactless technical approach to indoor human activity recognition (HAR). To address the difficulty of a single convolutional or recurrent network in simultaneously modeling local fluctuations, long-range temporal dependencies, and key action segments, this paper proposes CGAC, a model that integrates convolutional bidirectional gated recurrent units with temporal attention. The model first uses one-dimensional convolution and max pooling to extract and compress local temporal CSI features, then employs a BiGRU to model bidirectional contextual dependencies, and finally applies single-vector temporal attention to adaptively weight key time steps. Multi-dataset evaluations are conducted on three public datasets: UT-HAR, NTU-Fi HAR, and NTU-Fi Human-ID. CGAC achieves an accuracy of 99.70% on UT-HAR and accuracies of 97.50% and 97.81% on NTU-Fi HAR and NTU-Fi Human-ID, respectively. The results show that CGAC delivers the best performance on UT-HAR and remains competitive across different acquisition tools and CSI classification tasks.

An adaptive attention residual convolutional LSTM network for robust human activity recognition in healthy aging

LiteEvent-AE: Lightweight Autoencoder for Event-Based Vision on Low-Latency Energy-Constrained Edge Devices

A compact and configurable event-driven autoencoder that efficiently compresses neuromorphic data while preserving essential spatiotemporal structure for downstream inference and demonstrates the potential of compact event-driven models to advance environmentally conscious, low-power AI systems for high-speed perception in autonomous, mobile, and embedded computing environments.