Feature Identification With a Heuristic Algorithm and an Unsupervised Machine Learning Algorithm for Prior Knowledge of Gait Events.

Seth R Donahue,Michael E Hahn

doi:10.1109/tnsre.2021.3131953

Abstract

The purpose of this study was to compare a heuristic feature identification algorithm with output from the Beta Process Auto Regressive Hidden Markov Model (BP-AR-HMM) utilizing minimally sampled (≤ 100 Hz) human locomotion data for identification of gait events prior to their occurrence. Data were collected from 16 participants (21-64 years) using a single gyroscopic sensor in an inertial measurement unit on the dorsum of the foot, across multiple locomotion modes, including level ground walking and running (across speeds 0.8 m s-1 - 3.0 m s-1), ramps and stairs. Identification of gait events, initial contact (IC) and toe off (TO) with the heuristic algorithm, was 94% across locomotion modes. The features identified prior to initial contact had a lead time of 186.32 ± 86.70 ms, while TO had a lead time of 63.96 ± 46.30 ms. The BP-AR-HMM identified features that indicated an impending IC and TO with 99% accuracy, with a lead time of 59.41 ± 54.41 ms for IC and 90.79 ± 35.51 ms for TO. These approaches are consistent in their identification of gait events and have the potential to be utilized for classification and prediction of locomotion mode.

Highlights

HUMAN locomotion occurs across a range of speeds and environments that are not replicated in the laboratory
One example is switching linear dynamical systems, developed previously for economic analysis and the description of dynamical processes [33]. This class of algorithms allows for conditions where not all possible features are known a priori but can be added when a new feature is discovered. From this set of novel algorithms, we propose use of the Beta Process Auto Regressive Hidden Markov Model (BP-Auto-Regressive Hidden Markov Model (AR-HMM)) for the identification of gait events using a single stream of angular velocity data from a sensor on the foot sampled at 25 Hz
The heuristic algorithm was less accurate than the BP-AR-HMM the results remain similar to previous work with the addition of data from multiple locomotion modes collected in a real-world environment [3], [23]

Summary

Introduction

HUMAN locomotion occurs across a range of speeds and environments that are not replicated in the laboratory. Studies have utilized mobile sensor arrays for the quantification of human locomotion through varied environments [7] These mobile sensor arrays have allowed for the development of rules-based algorithms for the identification of gait events. Of particular interest, unsupervised machine learning algorithms for the analysis of multiple time series are being developed and tested for gait event detection [11], [12]. While these algorithms have been shown to be accurate for gait event identification in the laboratory, they have not been tested on data collected from a real-world environment, nor have they been compared to simple heuristic identifiers with minimal sensor data from the same data set

Objectives

Results

Discussion

Conclusion