Classification of ultra-short-term ECG samples: studies on events containing violence

Thesis event information

Date and time of the thesis defence

Place of the thesis defence

L10, https://oulu.zoom.us/j/63357450761

Topic of the dissertation

Classification of ultra-short-term ECG samples: studies on events containing violence

Doctoral candidate

Master of Science Hany Ferdinando

Faculty and unit

University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, OPEM

Subject of study

Electrical Engineering

Opponent

Professor Jukka Lekkala, Tampere University of Technology

Second opponent

Professor Heikki Ailisto, VTT

Custos

Professor Esko Alasaarela, University of Oulu

Add event to calendar

Identification of violent events from the ECG signal

Recent results from a study conducted at the University of Oulu confirm that the electronic signals of the heart (ECG signal) can be used to identify that a person has been subjected to violence. The research primarily aims at the automatic detection of school bullying. School violence is a serious problem with long-term consequences for the lives of up to one hundred million children around the world. The method developed in the study focuses on the victim’s perspective.

The results show that the method can identify changes in the ECG signal due to exposure to violence. The method makes it possible to issue an immediate alert to a school, guardian or social worker to help the victim. The study was conducted at the University of Oulu in collaboration with Petra Christian University (Indonesia), Harbin Institute of Technology (China) and Harbin University of Science and Technology (China). In Finland and Indonesia, several preliminary studies were conducted simulating students' violent behavior. These gathered valuable experience and information for the final simulation, which was attended by twelve 2nd and 5th grade students from a Chinese school in Harbin. They did various activities, some of which involved violence, some of which did not.

To process the results, several indices were quantified from the ECG signal, and by using different algorithms and machine learning methods we aimed to categorize, which activities contain violence and which do not. Detection accuracy was as high as 87%. Similar methods are used, for example, in computer games for biofeedback analysis and intelligent traffic systems. The development of a reliable and sufficiently accurate system for reporting incidents of violence requires further research. In order to improve the detection accuracy, other signals and data sources can be connected to the system using, for example, a camera, a microphone and motion sensors. There is a new glimmer of hope in sight for students suffering from school bullying.
Last updated: 3.12.2020