Towards Deep Clustering of Human Activities from Wearables

TL;DR

This paper introduces an unsupervised deep learning approach for human activity recognition using wearable sensor data, addressing the challenge of limited annotated datasets and demonstrating effective clustering of activities.

Contribution

It presents a novel end-to-end deep clustering method tailored for raw sequence data from wearables, advancing unsupervised activity recognition.

Findings

Effective joint learning of representations and clusters

Strong semantic correspondence with human activities

Outperforms existing unsupervised methods

Abstract

Our ability to exploit low-cost wearable sensing modalities for critical human behaviour and activity monitoring applications in health and wellness is reliant on supervised learning regimes; here, deep learning paradigms have proven extremely successful in learning activity representations from annotated data. However, the costly work of gathering and annotating sensory activity datasets is labor-intensive, time consuming and not scalable to large volumes of data. While existing unsupervised remedies of deep clustering leverage network architectures and optimization objectives that are tailored for static image datasets, deep architectures to uncover cluster structures from raw sequence data captured by on-body sensors remains largely unexplored. In this paper, we develop an unsupervised end-to-end learning strategy for the fundamental problem of human activity recognition (HAR) from wearables. Through extensive experiments, including comparisons with existing methods, we show the effectiveness of our approach to jointly learn unsupervised representations for sensory data and generate cluster assignments with strong semantic correspondence to distinct human activities.