2d08 formulation of machine learning model to classify human motion based on kinematics data of daily activities

Anuar, Mohamed (2022) 2d08 formulation of machine learning model to classify human motion based on kinematics data of daily activities. Faculty of Electrical and Electronic Engineering Technology, Universiti Malaysia Pahang.

EA18122_ANUAR_Thesis - Anuar Mohamed.pdf - Accepted Version

Download (2MB) | Preview


This project aims to formulate a method in "Machine Learning" to classify individual activities in the day based on daily data. "Machine Learning" is often used in today's technology. Malaysia has health problems caused by being overweight. Various campaigns were conducted to address this problem. Among them is the use of smart watches that can record a person's activities in the form of data. For example, this smart watch can be used in daily life to record how far (kilometers) a person walks or runs. The objective in the project is to find out the optimal location for the sensor on the individual body. In addition, it aims to identify the most appropriate method or model in Machine Learning to classify the activities of the individual. Next, it aims to investigate the accuracy in the classification. The project uses IMU-MMS as a sensor to record individual movement raw data. It will be placed on the wrist, back of body and shank as a location before the data is recorded. The recorded parameters are three axes (x, y, z) "accelerometer" and three axes (x, y, z) "gyroscope". The raw data will be then downloaded in csv file to the computer and then it will be analyzed in details such as Orange and Matlab. Methods in ML will be used such as Support Vector Machine, k-Nearest Neighbor, Neural Network and Naïve Bayes to train this data in classifying the type of movement of the individual. After that, tuning and analysis is done to get the best parameters for the data picture. After performing Analysis in Matlab and Orange by inserting 7200 data which consists of walking, running, standing, sitting, typing and jumping, overall, the wrist is the most optimum location for the IMU to record human motion or activites. "Neural Network" is the model in ML that provides the highest accuracy compared to other models in Orange. While kNN is the best model in Matlab. The accuracy achieved by most models is above 90%.

Item Type: Undergraduates Project Papers
Additional Information: SV: Dr. Nur Aqilah Binti Othman
Uncontrolled Keywords: human motion, kinematics data
Subjects: T Technology > TA Engineering (General). Civil engineering (General)
T Technology > TK Electrical engineering. Electronics Nuclear engineering
Faculty/Division: Faculty of Electrical and Electronic Engineering Technology
Depositing User: Mr. Nik Ahmad Nasyrun Nik Abd Malik
Date Deposited: 25 Oct 2023 03:11
Last Modified: 25 Oct 2023 03:11
URI: http://umpir.ump.edu.my/id/eprint/39014
Download Statistic: View Download Statistics

Actions (login required)

View Item View Item