2019

Sensors 2019, 19, 1820

Hidden Markov Model-Based Smart Annotation for Benchmark Cyclic Activity Recognition Database Using Wearables

Christine F. Martindale, Sebastijan Sprager, Bjoern M. Eskofier

Machine Learning and Data Analytics Lab, Computer Science Department, Erlangen

Keywords

activity recognition, benchmark database, gait analysis, inertial measurement unit, gait phases, cyclic activities, home monitoring, smart annotation, semi-supervised learning

Abstract

Activity monitoring using wearables is becoming ubiquitous, although accurate cycle level analysis, such as step-counting and gait analysis, are limited by a lack of realistic and labeled datasets. The effort required to obtain and annotate such datasets is massive, therefore we propose a smart annotation pipeline which reduces the number of events needing manual adjustment to 14%. For scenarios dominated by walking, this annotation effort is as low as 8%. The pipeline consists of three smart annotation approaches, namely edge detection of the pressure data, local cyclicity estimation, and iteratively trained hierarchical hidden Markov models. Using this pipeline, we have collected and labeled a dataset with over 150,000 labeled cycles, each with 2 phases, from 80 subjects, which we have made publicly available. The dataset consists of 12 different task-driven activities, 10 of which are cyclic. These activities include not only straight and steady-state motions, but also transitions, different ranges of bouts, and changing directions. Each participant wore 5 synchronized inertial measurement units (IMUs) on the wrists, shoes, and in a pocket, as well as pressure insoles and video. We believe that this dataset and smart annotation pipeline are a good basis for creating a benchmark dataset for validation of other semi- and unsupervised algorithms.

Moticon's Summary

In this study the authors introduce an annotation pipeline, which is intended to considerably reduce the effort for the manual event annotation in datasets of different movement activities obtained by sensor wearables. The proposed pipeline includes three annotation approaches i.e. edge detection of pressure data, local cyclicity estimation and iteratively trained hidden Markov models. The usability of the proposed annotation pipeline was demonstrated on a dataset for which data for different movement activities was collected from 80 subjects. Among other sensors Motcion sensor insoles were used for data collection. The authors were able to produce a dataset including multiple types of activities with over 150,000 labeled cycles with only 14% of events requiring manual adjustment. Furthermore, they believe that their annotation pipeline could aid in extending existing datasets.

Contact Us
Book a free online demo or use the contact form to get in touch
Newsletter
Subscribe to our newsletter for regular updates

Select your desired system

The cutting edge test based outcome assessment system for health professionals and trainers

The most versatile toolkit for free data acquisition and comprehensive analytics in research

Have a general inquiry?

Write us a message for general questions about products and solutions or if you’d like to discuss other topics.


The form was sent successfully.

You will be contacted shortly.

moticon-rego-sensor-insole-live-event

Stay one step ahead!

Subscribe to our newsletter for the latest information on case studies, webinars, product updates and company news

Get support

Check our FAQ database for answers to frequently asked questions

Describe your issue in as much detail as possible. Include screenshots or files if applicable.


Need help?
Want a live demo?
Interested in prices?
Want to say hello?
Always just a call away
+49 89 2000 301 60