Aims and scope

Neural Processing Letters is an international journal that promotes fast exchange of the current state-of-the art contributions among the artificial neural network community of researchers and users. The Journal publishes technical articles on various aspects of artificial neural networks and machine learning systems. Coverage includes novel architectures, supervised and unsupervised learning algorithms, deep nets, learning theory, network dynamics, self-organization, optimization, biological neural network modelling, and hybrid neural/fuzzy logic/genetic systems. The Journal publishes articles on methodological innovations for the applications of the afore mentioned systems in classification, pattern recognition, signal processing, image and video processing, robotics, control, autonomous vehicles, financial forecasting, big data analytics, and other multidisciplinary applications.

Special Issue on Machine Learning for Human Activity Recognition and Analysis
Submission Date: 2026-07-15

Human Activity Recognition (HAR) has emerged as a critical research domain at the intersection of sensing technologies, biomedical engineering, and artificial intelligence. By leveraging diverse signal modalities and advanced machine learning algorithms, HAR systems enable the analysis and interpretation of complex human behaviors across multiple application domains, including remote health monitoring, rehabilitation, assisted living, occupational safety, and adaptive human-computer interaction. Data-driven insights from HAR systems can inform clinical decision-making, optimize personalized health interventions, enhance elderly care, and improve user experiences in interactive technologies.

Recent advancements in sensing technologies and data collection methods—such as wearable sensors, mobile devices, and Internet of Things (IoT) systems—have revolutionized the HAR landscape. Modern systems now integrate multiple signal modalities, including inertial measurements (accelerometry, gyroscopy, magnetometry), physiological biosignals (ECG, EMG, EEG, PPG, respiration), acoustic and environmental signals, as well as video and radar data. When combined with sophisticated machine learning and deep learning techniques, these multimodal datasets facilitate accurate and context-aware recognition and analysis of human activities.

Despite rapid progress, challenges remain in developing comprehensive, well-annotated databases of physiological and behavioral signals. Ensuring data quality, annotation consistency, sensor placement standardization, and population diversity are essential for generalizable and reliable HAR models. Future advancements are expected to emphasize multimodal data fusion, explainable and interpretable AI models, edge computing for real-time processing, and privacy-preserving learning techniques such as federated learning. These innovations will be critical in enabling robust, adaptive, and personalized HAR systems.

This Collection aims to advance the state-of-the-art in HAR by exploring innovative approaches to signal processing, feature engineering, and machine learning. We welcome contributions addressing critical challenges, including:

- Identifying which signal modalities provide the most discriminative information for specific activity classes
- Developing multimodal fusion strategies to combine heterogeneous sensor streams
- Advancing deep learning architectures suited for temporal and spatial activity modeling
- Enhancing model interpretability, generalization, and energy efficiency
- Addressing real-world deployment constraints in clinical, industrial, and consumer applications

Topics of Interest Include (but are not limited to):

- Signal Modalities and Sensing Technologies: Inertial, physiological, acoustic, radar, and vision-based sensing
- Data Collection and Benchmarking: Sensing protocols, annotation methodologies, database construction, data quality assessment
- Signal Processing and Feature Engineering: Time-series analysis, wavelet transforms, feature selection, dimensionality reduction
- Multimodal Data Fusion: Early, late, and hybrid fusion strategies; attention-based fusion mechanisms
- Machine Learning and Deep Learning: CNNs, RNNs, LSTMs, Transformers, GNNs, attention mechanisms for activity recognition
- Transfer Learning and Domain Adaptation: Cross-subject, cross-dataset, and cross-device generalization
- Real-World Implementation: Edge AI, energy-efficient models, privacy-preserving and federated learning techniques
- Clinical and Healthcare Applications: Fall detection, gait and posture analysis, rehabilitation assessment, mental health monitoring

This Collection supports and amplifies research related to- SDG 11: Sustainable Cities & Communities; SDG 9: Industry, Innovation & Infrastructure; SDG 3: Good Health & Well-Being

We invite researchers to submit their work to this Collection, which will showcase innovative research addressing these and related topics in machine learning-driven human activity recognition and analysis.