Artificial intelligence based analysis and game control using brain signals

• Main Author: Hong, Calvin Wen Kit
• Other Authors: Smitha Kavallur Pisharath Gopi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2022
• Subjects: Science::Medicine::Biosensors; Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/156470
• Institution: Nanyang Technological University

Brain Computer Interface is an emerging field which is gaining traction particularly in the use for video games. This is done by extracting brain signals via a conventional method used in the biomedical domain, which is Electroencephalography(EEG). The project involves developing the entire pipeline of the signal extraction, processing, classification, game development and interfacing. A low cost, low electrode count EEG capturing device, known as the InterAxon Muse 2, is used as the choice of hardware for this project. This project also aims to evaluate the feasibility of using the InterAxon Muse 2 as the EEG capturing choice. In particular, this project has two main points to evaluate; the first is to compare and contrast the separability of attentive and relax states, and the second is to evaluate whether the model is transferable onto a game. Experiments were conducted using various neurocognitive prompts such as the flanker task and the proposed majority task with varying block length. The results are visualized using Principle Component Analysis and are trained on SVM and Neural Network models. A simple ball game, created using Unity3D is used in tandem with the trained models to evaluate the transferability of the model for the game. Results reflect that Neural Networks have a higher performance as compared to SVMs, and that the proposed majority task has a higher classification accuracy based on the mean obtained from the 10 subjects, and that the models are generally transferable onto the ball game created.