Oscillatory Source Tensor Discriminant Analysis (OSTDA): A regularized tensor pipeline for SSVEP-based BCI systems

• BCI pipeline with tensor feature reduction and analytical regularization. Comparison three state-of-the-art methods during different perturbations. Achieves high accuracy in short trials small sample sizes. Offers performance both low number of electrodes. Suitable for noisy data, channel setups limited training data. Periodic signals called Steady-State Visual Evoked Potentials (SSVEP) are elicited the brain by flickering stimuli. They usually detected means regression techniques that need relatively long trial lengths to provide feedback and/or sufficient calibration be reliably estimated context brain-computer interface (BCI). Thus, systems designed operate SSVEP signals, reliability is achieved at expense speed or extra recording time. Furthermore, regardless length, free regression-based have been shown suffer from significant drops when cognitive perturbations present affecting attention In this study we a novel technique Oscillatory Source Tensor Discriminant Analysis (OSTDA) extracts oscillatory sources classifies them using newly developed tensor-based discriminant analysis shrinkage. The proposed approach robust size settings where only few available. Besides, it works well low- high-number-of-channel settings, as one second. OSTDA performs similarly significantly better than other benchmarked under experimental including those disturbances (i.e. four datasets control, listening, speaking thinking conditions). Overall, paper show among all studied ones can achieve optimal results analyzed conditions.