Active Machine Learning For Heterogeneity Activity Recognition through Smartwatch Sensors

Abstract

The proliferation of wearable technology, especially smartwatches, has furnished a rich source of statistics for hobby popularity. This paper explores the utility of lively device mastering techniques to apprehend heterogeneous activities via smartwatch sensors. The proposed gadget leverages accelerometer and gyroscope information to categorise numerous physical activities inclusive of on foot, strolling, biking, and standing, amongst others. Active mastering is utilized to beautify the version's performance with the aid of selectively querying the maximum informative statistics points, as a result minimizing the number of categorized facts required. The methodology entails initial schooling with a small categorised dataset, followed via iterative cycles of lively getting to know to refine the version. Experimental results display that the proposed technique achieves excessive accuracy and robustness in hobby reputation, outperforming traditional gadget learning techniques. This study underscores the potential of lively mastering in decreasing the labeling effort at the same time as keeping high category accuracy, making it a feasible solution for actual-time hobby monitoring in ubiquitous healthcare and health packages.