PRICAI 2019: Trends in Artificial Intelligence

Abstract

Over the last decade, processing of biomedical signals using machine learning algorithms has gained widespread attention. Amongst these, one of the most important signals is electroencephalography (EEG) signal that is used to monitor the brain activities. Brain-computer-interface (BCI) has also become a hot topic of research where EEG signals are usually acquired using non-invasive sensors. In this work, we propose a scheme based on common spatial spectral pattern (CSSP) and optimization of temporal filters for improved motor imagery (MI) EEG signal recognition. CSSP is proposed as it improves the spatial resolution while the temporal filter is optimized for each subject as the frequency band which contains most significant information varies amongst different subjects. The proposed scheme is evaluated using two publicly available datasets: BCI competition III dataset IVa and BCI competition IV dataset 1. The proposed scheme obtained promising results and outperformed other state-of-the-art methods. The findings of this work will be beneficial for developing improved BCI systems.