A novel WGF-LN based edge driven intelligence for wearable devices in human activity recognition

Area of Science:

• Health monitoring
• Machine Learning
• Signal Processing

Background:

• Human Activity Recognition (HAR) relies on wearable sensors for continuous data streams.
• Supervised machine learning (ML) and deep learning (DL) show promise but face decision-making ambiguity.
• Existing methods struggle with the complexity and ambiguity of sensor data for HAR.

Purpose of the Study:

• To propose a novel Wasserstein gradient flow legonet (WGF-LN) system for improved HAR.
• To address the challenges of decision-making ambiguity in sensor data analysis for HAR.
• To enhance the accuracy and efficiency of human activity classification using wearable sensor data.

Main Methods:

• Data pre-processing followed by Haar Wavelet mother-Symlet wavelet coefficient scattering feature extraction (HS-WSFE).
• Feature selection using Binomial Distribution integrated-Golden Eagle Optimization (BD-GEO).
• Post-processing of features using scatter plot matrix and classification with WGF-LN.

Main Results:

• The proposed WGF-LN system demonstrated significant efficacy in human activity recognition.
• The integrated feature extraction, selection, and classification pipeline yielded accurate results.
• Experimental validation confirmed the effectiveness of the novel HAR system.

Conclusions:

• The developed WGF-LN system offers a robust solution for HAR.
• The proposed methodology effectively handles sensor data complexity and reduces ambiguity.
• This research contributes a promising advancement in wearable health monitoring and activity classification.