Abstract:
Brain waves contain fundamental information about cortical activity: signal power within certain frequency bands, which is exploited by a variety of Brain-Computer Interf...View moreMetadata
Abstract:
Brain waves contain fundamental information about cortical activity: signal power within certain frequency bands, which is exploited by a variety of Brain-Computer Interface applications. For real-time systems, these features must be estimated as quickly as possible while maintaining high signal fidelity. Here, we present a statistically optimal signal processing framework for real-time bandpower estimation and tracking. Key components are a spectral shaping stage for increased sensitivity and Kalman filtering of log-transformed bandpower estimates for optimal tracking. The system has one degree of freedom, which allows for adaptive design based on signal dynamics. The overall complexity remains low. We evaluated the proposed architecture based on two experiments involving cortical motor functions and receptive-language related cortical areas. First results are promising. Spectral shaping based on a whitening transform increases the sensitivity (z-Score) by up to 60 %. Furthermore, the tracking time lag is substantially reduced relative to conventional approaches.
Date of Conference: 25-28 May 2017
Date Added to IEEE Xplore: 14 August 2017
ISBN Information:
Electronic ISSN: 1948-3554