Optimising Systems Combining Sensor Inputs: An EEG-Based Brain-Computer Interface Game As A Case Study

Abstract

With developments in the field of sensor technology, devices have become increasing portable, and the processing of signals from these sensors is often outsourced to an additional device away from the sensor itself. When performing computation locally, processing is usually more costly in terms of both energy and time; however sending it to a remote device would be costly in terms of communication energy. This thesis aims to investigate these trade-offs in energy over a tiered network of computation, in order to implement an optimised system. The system that will be optimised in this thesis is a brain-computer interface that uses two live streams of EEG data to control a two-player game. This report first describes the work involved in characterising the EEG sensor and implementing the brain-computer interface system. It then goes on to discuss the energy trade-offs in the system and in alternative system configurations. A different application is also profiled - an ECG arrhythmia detector. The findings from profiling these two systems are used to discuss energy trade-offs in the wider sense and how energy estimates could be made in other potential applications with distributed computation.