An ARM Cortex Microcontroller Based Solution for Real-Time Extraction and Classification of Autoregressive EEG Features

Abstract

Recent years have seen a boom in the use of Electroencephalography (EEG) to catch brain waves because of its high temporal resolution, non-invasive nature, and affordability. However, most of the EEG processing solutions are based on computers or proprietary ASICs. This paper presents a low-cost general-purpose microcontroller based system that can extract and classify EEG features in real-time (up to 40k samples/sec/channel) for the purpose of controlling the movement of a robot in two directions: left and right. The microcontroller employs Recursive Least Squares (RLS) algorithm to extract the autoregressive (AR) features from EEG signals, and then classifies them using Linear Discriminant Analysis (LDA) classifier. A microcontroller implementation has various advantages over computer based systems: reduced power consumption, weight and cost. Also, availability of low-level I/O controls make it possible to handle sensors and actuators without the need of additional hardware. These advantages are most prominent in embedded systems (such as wheel chairs, powered prosthesis etc.) where limited energy is available to power the processor and carry weights.