Enhancing Temporal Classification of AAR Parameters in EEG single-trial analysis for Brain-Computer Interfacing

Adaptive autoregressive (AAR) coefficients provide dynamic spectral information in EEG single-trial analysis. In this paper we propose a temporal evidence accumulation framework to enhance classification of AAR features. The results for a single subject, using 280 trials, indicate distinct improvements over a conventional method of temporal classification. We illustrate how the framework is applicable to AAR features, as well as to wavelet features as reported in Lemm et al., (2004). These findings put the two time-frequency features on equal footing for comparison in this context