Multi-Frequency Federated Learning for Human Activity Recognition Using Head-Worn Sensors

TL;DR

This paper introduces a multi-frequency federated learning approach for human activity recognition using head-worn sensors, enhancing privacy and model learning across devices with different sampling rates.

Contribution

It proposes a novel multi-frequency federated learning framework tailored for head-worn devices, addressing privacy concerns and enabling joint model training across diverse sensor sampling frequencies.

Findings

Improved accuracy over frequency-specific methods on two datasets

Demonstrated feasibility of multi-frequency FL in head-worn HAR

Publicly available implementation for further research

Abstract

Human Activity Recognition (HAR) benefits various application domains, including health and elderly care. Traditional HAR involves constructing pipelines reliant on centralized user data, which can pose privacy concerns as they necessitate the uploading of user data to a centralized server. This work proposes multi-frequency Federated Learning (FL) to enable: (1) privacy-aware ML; (2) joint ML model learning across devices with varying sampling frequency. We focus on head-worn devices (e.g., earbuds and smart glasses), a relatively unexplored domain compared to traditional smartwatch- or smartphone-based HAR. Results have shown improvements on two datasets against frequency-specific approaches, indicating a promising future in the multi-frequency FL-HAR task. The proposed network's implementation is publicly available for further research and development.