LaHAR: Latent Human Activity Recognition using LDA

TL;DR

This paper introduces LaHAR, a novel method using Latent Dirichlet Allocation to discover latent human activity patterns from sequential sensor data, providing an interpretable and effective clustering approach for HAR.

Contribution

It applies LDA, originally for text analysis, to sensor data for the first time in HAR, enabling discovery of underlying activity patterns without predefined classes.

Findings

LDA effectively uncovers latent structures in HAR data.

LDA achieves accurate clustering of activity sequences.

Provides human-understandable representations of sensor data.

Abstract

Processing sequential multi-sensor data becomes important in many tasks due to the dramatic increase in the availability of sensors that can acquire sequential data over time. Human Activity Recognition (HAR) is one of the fields which are actively benefiting from this availability. Unlike most of the approaches addressing HAR by considering predefined activity classes, this paper proposes a novel approach to discover the latent HAR patterns in sequential data. To this end, we employed Latent Dirichlet Allocation (LDA), which is initially a topic modelling approach used in text analysis. To make the data suitable for LDA, we extract the so-called "sensory words" from the sequential data. We carried out experiments on a challenging HAR dataset, demonstrating that LDA is capable of uncovering underlying structures in sequential data, which provide a human-understandable representation of the data. The extrinsic evaluations reveal that LDA is capable of accurately clustering HAR data sequences compared to the labelled activities.