Daily Physical Activity Monitoring -- Adaptive Learning from Multi-source Motion Sensor Data

TL;DR

The paper tackles reliable daily physical activity monitoring with a single wearable by transferring knowledge from lab-based multi-source sensor data. It introduces Inter-domain Pairwise Distance ($IPD$), a metric that leverages the pairwise structure of multi-source time series to quantify domain similarity and guide transfer learning. The method pre-trains a unified classifier on source domains with IPD-guided adaptive learning rates, then fine-tunes on the target domain, and demonstrates superior accuracy and robustness to noise on the UCI DSA dataset and an RSS-based dataset. This approach reduces the performance gap between lab-rich multi-sensor data and everyday single-sensor use, offering practical improvements for continuous health monitoring and early intervention. The framework’s emphasis on domain-pairing structure and smooth bootstrap-based IPD estimation provides a robust, generalizable pathway for transferring laboratory insights to real-world wearable sensing applications.

Abstract

In healthcare applications, there is a growing need to develop machine learning models that use data from a single source, such as that from a wrist wearable device, to monitor physical activities, assess health risks, and provide immediate health recommendations or interventions. However, the limitation of using single-source data often compromises the model's accuracy, as it fails to capture the full scope of human activities. While a more comprehensive dataset can be gathered in a lab setting using multiple sensors attached to various body parts, this approach is not practical for everyday use due to the impracticality of wearing multiple sensors. To address this challenge, we introduce a transfer learning framework that optimizes machine learning models for everyday applications by leveraging multi-source data collected in a laboratory setting. We introduce a novel metric to leverage the inherent relationship between these multiple data sources, as they are all paired to capture aspects of the same physical activity. Through numerical experiments, our framework outperforms existing methods in classification accuracy and robustness to noise, offering a promising avenue for the enhancement of daily activity monitoring.

Figures (3)

• Figure 1: The procedure of training a classifier and applying the trained classifier in daily physical activity monitoring.
• Figure 2: The illustration of the proposed transfer learning framework.
• Figure 3: RCC of time series classification approaches across different ratios of noise. The standard deviation of RCC is indicated by the shadow along the line.

Theorems & Definitions (4)

• Definition 1: Time series data
• Definition 2: Domain
• Definition 3: Pairwise multi-source time series
• Definition 4