Brain Computer Interfacing With Optimized Output Encoding

This page is dedicated to my Master Thesis Brain Computer Interfacing With Optimized Output Encoding which was part of my MSc in Biomedical Engineering at the University of Bern.

Introduction

This thesis concludes the efforts to investigate the combination of current Brain Computer Interfaces (BCI) systems, which mostly follow a machine learning and pattern recognition approach to do classification, and Statistical Language Modeling (SLM), to improve classification rate in current BCI systems.

BCI systems today use a wide variety of techniques for classification, based, for instance, on dynamic linear regression models, Kalman filters, hidden Markov models, and neural networks, to name only few of them. These techniques all have the same goal: to improve BCI performance by either faster and more accurate classification, or by improving the number of mental tasks that can be classified.

Statistical language modeling (SLM) is an attempt to capture regularities of natural language for the purpose of improving the performance of various natural language applications. Statistical language modeling is crucial for a large variety of language technology applications. These include speech recognition, machine translation, document classification and routing, optical character recognition, information retrieval, handwriting recognition, spelling correction, genomics, as well as many more. The most successful SLM techniques use very little knowledge of what language really is. The most popular language models (n-grams) take no advantage of the fact that what is being modeled is language – it may as well be a sequence of arbitrary symbols!

The combination of the two fields is the work concluded in this thesis and was expected to lead to a more robust BCI system. The idea was to use the output of a well performing and fast two-class classifier in combination with a graphical user interface, which when backed by a statistical language model would allow information efficient state selection, so gaining an increased information transfer rate and accuracy in BCI applications. Files

> Get the thesis here.