Data Reconstruction Methods in Multi-Feature Fusion CNN Model for Enhanced Human Activity Recognition

Sensors, Journal Year: 2025, Volume and Issue: 25(4), P. 1184 - 1184

Published: Feb. 14, 2025

Human activity recognition (HAR) plays a pivotal role in digital healthcare, enabling applications such as exercise monitoring and elderly care. However, traditional HAR methods relying on accelerometer data often require complex preprocessing steps, including noise reduction manual feature extraction. Deep learning-based human using one-dimensional suffers from limited Transforming time-series signals into two-dimensional representations has shown potential for enhancing extraction reducing noise. existing single-feature inputs or extensive face limitations robustness accuracy. This study proposes multi-input, CNN architecture three distinct reconstruction methods. By fusing features reconstructed images, the model enhances capabilities. method was validated custom dataset without requiring steps. The proposed outperformed models single-reconstruction raw data. Compared to baseline, it achieved 16.64%, 13.53%, 16.3% improvements accuracy, precision, recall, respectively. We tested across various levels of noise, consistently demonstrated greater than time-series-based approach. Fusing effectively captured latent patterns variations demonstrates that can be improved multi-input approach with offers practical efficient solution, streamlining performance, making suitable real-world applications.

Language: Английский

A machine-learning approach for stress detection using wearable sensors in free-living environments

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 179, P. 108918 - 108918

Published: July 18, 2024

Language: Английский

NeuroHAR: A Neuroevolutionary Method for Human Activity Recognition (HAR) for Health Monitoring

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 112232 - 112248

Published: Jan. 1, 2024

Language: Английский