Logo

A Combined Motion-Audio School Bullying Detection Algorithm

Ye, Liang and Ferdinando, Hany and Wang, Peng and Wang, Le and Seppänen, Tapio and ALASAARELA, ESKO (2018) A Combined Motion-Audio School Bullying Detection Algorithm. [UNSPECIFIED]

[img] PDF
Download (480Kb)

    Abstract

    School bullying is a common social problem which affects children both mentally and physically, and preventing school bullying is a timeless topic all over the world. This paper proposes a school bullying detection method based on activity recognition and speech emotion recognition. Motion data and voice data are gathered by movement sensors and microphone, and motion features and audio features are extracted to describe bullying events and daily-life events. Motion features include time domain features and frequency domain features. Audio features are the classical MFCCs. Wrapper is used for feature selection. Then motion features and audio features together form combined feature vectors for classification, and LDA is used for further dimension reduction. A BPNN is trained to recognize bullying activities and distinguish them from daily-life ones. An action transition detection method is proposed to reduce computational complex for the purpose of practical use. Only when an action transition event has been detected, the school bullying detection algorithm will run. Simulation results show that the motion-audio combined feature vector outperforms sole motion features and sole acoustic features, with accuracy of 82.4% and precision of 92.2%. Moreover, with the action transition method, the computation can be reduced by half.

    Item Type: UNSPECIFIED
    Uncontrolled Keywords: activity recognition; speech emotion recognition; movement sensors; school bullying; pattern recognition
    Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
    Q Science > QA Mathematics > QA76 Computer software
    Divisions: Faculty of Industrial Technology > Electrical Engineering Department
    Depositing User: Admin
    Date Deposited: 09 Sep 2018 19:46
    Last Modified: 16 Jul 2019 10:57
    URI: http://repository.petra.ac.id/id/eprint/17950

    Actions (login required)

    View Item