Abstract:Wearable sensor-based human activity recognition (HAR) is a critical research domain in activity perception. However, achieving high efficiency and long sequence recognition remains a challenge. Despite the extensive investigation of temporal deep learning models, such as CNNs, RNNs, and transformers, their extensive parameters often pose significant computational and memory constraints, rendering them less suitable for resource-constrained mobile health applications. This study introduces HARMamba, an innovative light-weight and versatile HAR architecture that combines selective bidirectional State Spaces Model and hardware-aware design. To optimize real-time resource consumption in practical scenarios, HARMamba employs linear recursive mechanisms and parameter discretization, allowing it to selectively focus on relevant input sequences while efficiently fusing scan and recompute operations. The model employs independent channels to process sensor data streams, dividing each channel into patches and appending classification tokens to the end of the sequence. It utilizes position embedding to represent the sequence order. The patch sequence is subsequently processed by HARMamba Block, and the classification head finally outputs the activity category. The HARMamba Block serves as the fundamental component of the HARMamba architecture, enabling the effective capture of more discriminative activity sequence features. HARMamba outperforms contemporary state-of-the-art frameworks, delivering comparable or better accuracy with significantly reducing computational and memory demands. It's effectiveness has been extensively validated on 4 publically available datasets namely PAMAP2, WISDM, UNIMIB SHAR and UCI. The F1 scores of HARMamba on the four datasets are 99.74%, 99.20%, 88.23% and 97.01%, respectively.

Markov Dynamic Subsequence Ensemble for Energy-Efficient Activity Recognition.

Wearable Sensor Based Multimodal Human Activity Recognition Exploiting the Diversity of Classifier Ensemble.

Learning Datum-Wise Sampling Frequency for Energy-Efficient Human Activity Recognition

Sensor-Based Human Activity Mining Using Dirichlet Process Mixtures of Directional Statistical Models

EMSN: an Energy-Efficient Memristive Sequencer Network for Human Emotion Classification in Mental Health Monitoring

Energy Expenditure Estimation Through Daily Activity Recognition Using a Smart-phone

Ensembles of Deep LSTM Learners for Activity Recognition using Wearables

Human Activity Recognition from Smart-Phone Sensor Data using a Multi-Class Ensemble Learning in Home Monitoring

EnFed: An Energy-aware Opportunistic Federated Learning in Resource Constrained Environments for Human Activity Recognition

MMTSA: Multimodal Temporal Segment Attention Network for Efficient Human Activity Recognition

Seeker: Synergizing Mobile and Energy Harvesting Wearable Sensors for Human Activity Recognition

Motion Sequence Analysis Using Adaptive Coding with Ensemble Hidden Markov Models

Ensemble of deep learning techniques to human activity recognition using smart phone signals

An Energy-Efficient Ensemble Approach for Mitigating Data Incompleteness in IoT Applications

Uncertainty Quantification for Deep Context-Aware Mobile Activity Recognition and Unknown Context Discovery

randomHAR: Improving Ensemble Deep Learners for Human Activity Recognition with Sensor Selection and Reinforcement Learning

Feedback-based metric learning for activity recognition

Instance-Wise Dynamic Sensor Selection for Human Activity Recognition

HARMamba: Efficient and Lightweight Wearable Sensor Human Activity Recognition Based on Bidirectional Mamba

Learning Human Activity Containing Sparse Irrelevant Events in Long Sequence

Employment of Ensemble Machine Learning Methods for Human Activity Recognition