University of Oulu

Detecting head movement using gyroscope data collected via in-ear wearables

Saved in:
Author: Islam, Adnanul1
Organizations: 1University of Oulu, Faculty of Information Technology and Electrical Engineering, Computer Science
Format: ebook
Version: published version
Access: open
Online Access: PDF Full Text (PDF, 2.9 MB)
Pages: 71
Persistent link:
Language: English
Published: Oulu : A. Islam, 2021
Publish Date: 2021-06-23
Thesis type: Master's thesis (tech)
Tutor: Peltonen, Ella
Reviewer: Pirttikangas, Susanna
Peltonen, Ella


Head movement is considered as an effective, natural, and simple method to determine the pointing towards an object. Head movement detection technology has significant potentiality in diverse field of applications and studies in this field verify such claim. The application includes fields like users interaction with computers, controlling many devices externally, power wheelchair operation, detecting drivers’ drowsiness while they drive, video surveillance system, and many more. Due to the diversity in application, the method of detecting head movement is also wide-ranging. A number of approaches such as acoustic-based, video-based, computer-vision based, inertial sensor data based head movement detection methods have been introduced by researchers over the years. In order to generate inertial sensor data, various types of wearables are available for example wrist band, smart watch, head-mounted device, and so on.

For this thesis, eSense — a representative earable device — that has built-in inertial sensor to generate gyroscope data is employed. This eSense device is a True Wireless Stereo (TWS) earbud. It is augmented with some key equipment such as a 6-axis inertial motion unit, a microphone, and dual mode Bluetooth (Bluetooth Classic and Bluetooth Low Energy). Features are extracted from gyroscope data collected via eSense device. Subsequently, four machine learning models — Random Forest (RF), Support Vector Machine (SVM), Naïve Bayes, and Perceptron — are applied aiming to detect head movement. The performance of these models is evaluated by four different evaluation metrics such as Accuracy, Precision, Recall, and F1 score. Result shows that machine learning models that have been applied in this thesis are able to detect head movement. Comparing the performance of all these machine learning models, Random Forest performs better than others, it is able to detect head movement with approximately 77% accuracy. The accuracy rate of other three models such as Support Vector Machine, Naïve Bayes, and Perceptron is close to each other, where these models detect head movement with about 42%, 40%, and 39% accuracy, respectively. Besides, the result of other evaluation metrics like Precision, Recall, and F1 score verifies that using these machine learning models, different head direction such as left, right, or straight can be detected.

see all

Copyright information: © Adnanul Islam, 2021. This publication is copyrighted. You may download, display and print it for your own personal use. Commercial use is prohibited.