Estimating Change Intensity and Duration in Human Activity Recognition using Martingales

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The subject of physical activity is becoming prominent in the healthcare system for the improvement and monitoring of movement disabilities. Existing algorithms are effective in detecting changes in data streams but most of these approaches are not focused on measuring the change intensity and duration. In this paper, we improve on the geometric moving average martingale method by optimising the parameters in the weighted average using a genetic algorithm. The proposed approach enables us to estimate the intensity and duration of transitions that happen in human activity recognition scenarios. Results show that the proposed method makes some improvement over previous martingale techniques.
Original languageEnglish
Title of host publicationThe 7th International Online & Onsite Conference on Machine Learning, Optimization, and Data Science
Subtitle of host publicationLOD2021
PublisherSpringer
Publication statusAccepted/In press - 31 Jul 2021
EventThe 7th International Online & Onsite Conference on Machine Learning, Optimization, and Data Science - Grasmere England, Lake Districk, United Kingdom
Duration: 4 Oct 20218 Oct 2021
https://lod2021.icas.cc/

Conference

ConferenceThe 7th International Online & Onsite Conference on Machine Learning, Optimization, and Data Science
Abbreviated titleLOD2021
CountryUnited Kingdom
CityLake Districk
Period4/10/218/10/21
Internet address

Keywords

  • Change detection
  • martingales
  • human activity recognition

Fingerprint

Dive into the research topics of 'Estimating Change Intensity and Duration in Human Activity Recognition using Martingales'. Together they form a unique fingerprint.

Cite this