Multivariate Time Series Classification Using Dynamic Time Warping Template Selection for Human Activity Recognition

TL;DR

This paper introduces a template selection method using Dynamic Time Warping for human activity recognition, reducing the need for complex feature extraction and domain knowledge, and demonstrating effectiveness on simulated and real smartphone data.

Contribution

It presents a novel template selection approach with DTW for activity classification that simplifies the process by avoiding complex feature engineering.

Findings

Effective classification on simulated data

Successful application to real smartphone data

Reduces reliance on feature engineering

Abstract

Accurate and computationally efficient means for classifying human activities have been the subject of extensive research efforts. Most current research focuses on extracting complex features to achieve high classification accuracy. We propose a template selection approach based on Dynamic Time Warping, such that complex feature extraction and domain knowledge is avoided. We demonstrate the predictive capability of the algorithm on both simulated and real smartphone data.