Deep Learning-Based Recognition of Human Activities From Wearable Sensor Data

Budi Santoso

Deep Learning-Based Recognition of Human Activities From Wearable Sensor Data

Author(s)	Budi Santoso
Country	Indonesia
Abstract	Human activity recognition has become an important research area within artificial intelligence, wearable computing, and intelligent sensor-based information systems. Wearable devices equipped with accelerometers, gyroscopes, and related motion sensors generate continuous measurements that can be analyzed to identify different patterns of human movement. These observations provide opportunities for automated classification of activities such as walking, stair climbing, sitting, standing, and lying. However, activity recognition remains challenging because human movements exhibit considerable variation in speed, duration, intensity, sensor orientation, and individual performance characteristics. Conventional machine learning approaches frequently depend on manually constructed features that may not adequately capture complex temporal relationships within wearable sensor signals. The present paper develops a hypothetical methodological framework for examining deep learning-based human activity recognition using multichannel wearable sensor data. The proposed framework compares Random Forest classification using engineered features, a one-dimensional Convolutional Neural Network, a Bidirectional Long Short-Term Memory network, and a hybrid CNN–LSTM architecture. A fictional dataset containing 15,000 sensor observation windows associated with 60 hypothetical participants is introduced to demonstrate the intended research design. Six everyday activities are considered as classification categories. The proposed methodology incorporates sensor signal preprocessing, temporal segmentation, feature representation, neural network development, participant-independent validation, and comparative classification assessment. Two illustrative analytical tables describe the hypothetical activity distribution and comparative classification performance. Accuracy and macro-averaged F1-score are considered principal evaluation measures. All numerical values are manually constructed and do not represent actual sensor recordings, executed experiments, or independently verified classification outcomes. The methodological discussion examines sensor placement, temporal dependencies, activity transitions, interindividual variation, computational requirements, and model generalization. Particular attention is given to the importance of participant-level data separation and the distinction between laboratory-based recognition accuracy and practical performance under everyday conditions. The paper concludes that deep learning provides a relevant methodological foundation for wearable activity recognition, although its effectiveness must be established through representative datasets, reproducible experiments, and independent evaluation.
Keywords	Human Activity Recognition, Deep Learning, Wearable Sensors, Convolutional Neural Networks, Long Short-Term Memory, Accelerometer, Gyroscope, Sensor Data Classification, Activity Monitoring, Artificial Intelligence.
Field	Engineering
Published In	Volume 2, Issue 1, January-February 2026
Published On	2026-01-02

View / Download PDF File

E-ISSN: 3067-5073

doi

CrossRef DOI is assigned to each research paper published in our journal.

IJMLAIDSE DOI prefix is
10.00000/IJMLAIDSE

Downloads

Research Paper Format Copyright Permission and Undertaking Form

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About IJMLAIDSE Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us	+1-872-302-3851	editor@uscse.com

International Journal of Machine Learning, AI & Data Science Evolution

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Monthly Scholarly International Journal

Deep Learning-Based Recognition of Human Activities From Wearable Sensor Data

Share this