Implement an autocomplete function for a VEP BCI Speller

Bachelor’s / Master’s Thesis, assigned to Ulzii-Utas Narantsatsralt
Master's student
Lab: Neural Interfaces and Brain Signal Decoding
Ulzii-Utas Narantsatsralt

Abstract

A Brain-Computer Interface (BCI) is an interface between the human brain and a computer. BCIs enable physically paralyzed patients, for example, to spell via pure brain activity. The EEG2Code BCI developed in Tübingen is currently the fastest non-invasive BCI system. While the classification accuracy is around 100%, the letter-by-letter communication speed has reached its maximum due to the human response time. In order to improve the communication speed even further, the user must train accordingly to become faster. Another possibility is to use a language model, such as the one known from mobile phones. After typing one (or more) characters, words are suggested which are most likely to be heard, and the suggestions adapt to the user’s typing behaviour over time. Such an auto-completion can increase the writing speed even more, especially for words with many characters.

Task

The aim of this work is to implement such a autocomplete function. Depending on whether it is a Bachelor’s or Master’s thesis, different requirements are set. These range from a simple word frequency to a complex natural language processing (NLP) method. Details are discussed in a first meeting and can be adjusted according to the progress made. As a proof on concept, it is desired to test the final BCI system on itself.

Requirements

  • Interest in neuroscience
  • Good programming skills
  • Experience in MATLAB is beneficial

Contact

Nagel, Sebastian