Left-Right Swapping and Upper-Lower Limb Pairing for Robust Multi-Wearable Workout Activity Detection

TL;DR

This paper introduces novel data augmentation techniques for multi-wearable activity detection, improving model accuracy by addressing wearable orientation inconsistencies in accelerometer data.

Contribution

It proposes and evaluates left-right swapping and upper-lower limb pairing augmentation methods for robust activity classification with multiple wearables.

Findings

Left-right swapping improves F1-score by 1.29%.

Upper-lower limb pairing further enhances accuracy to 91.87%.

Traditional machine learning models benefit from these augmentation techniques.

Abstract

This work presents the solution of the Signal Sleuths team for the 2024 HASCA WEAR challenge. The challenge focuses on detecting 18 workout activities (and the null class) using accelerometer data from 4 wearables - one worn on each limb. Data analysis revealed inconsistencies in wearable orientation within and across participants, leading to exploring novel multi-wearable data augmentation techniques. We investigate three models using a fixed feature set: (i) "raw": using all data as is, (ii) "left-right swapping": augmenting data by swapping left and right limb pairs, and (iii) "upper-lower limb paring": stacking data by using upper-lower limb pair combinations (2 wearables). Our experiments utilize traditional machine learning with multi-window feature extraction and temporal smoothing. Using 3-fold cross-validation, the raw model achieves a macro F1-score of 90.01%, whereas left-right swapping and upper-lower limb paring improve the scores to 91.30% and 91.87% respectively.