[home] [Personal Program] [Help]

---

Session: Poster session I
Session starts: Thursday 24 January, 15:00

---

Simon Van Eyndhoven (KU Leuven)
Martijn Boussé (KU Leuven)
Borbála Hunyadi ()
Lieven De Lathauwer (KU Leuven Kulak)
Sabine Van Huffel (KU Leuven)

Abstract:
The classification of brain states using neural recordings such as electroencephalography (EEG) finds applications in both medical and non-medical contexts, such as detecting epileptic seizures or discriminating mental states in brain-computer interfaces, respectively. Although this endeavour is well-established, existing solutions are typically restricted to lab or hospital conditions because they operate on recordings from a set of EEG electrodes that covers the whole head. By contrast, a true breakthrough for these applications would be the deployment ‘in the real world’, by means of wearable devices that encompass just one (or a few) channels. Such a reduction of the available information inevitably makes the classification task more challenging. To address this issue, we developed a pipeline for classifying brain states based on data from only a single EEG channel, after a calibration phase in which information of multiple channels is exploited. The pipeline relies on 1) a multilinear subspace learning step, in which spectral filters are tuned that extract frequency bands that are discriminative for the classification task, 2) solving a tensor regression problem with a low-rank structure and 3) the application of an off-the-shelf classifier on the estimated regression coefficients. We demonstrate the feasibility of this approach for EEG data recorded during a mental arithmetic task. However, it is generic; it may be applicable for the detection of e.g. epileptic seizures as well, as these also induce spectral changes, and for non-EEG data. The proposed framework can readily be improved, by using a few channels instead of only one, and by rigorously tuning the parameters in the pipeline with cross-validation. [1] Van Eyndhoven S., Bousse M., Hunyadi B., De Lathauwer L., Van Huffel S., “Single-channel EEG classification by multi-channel tensor subspace learning and regression”, ESAT-STADIUS, KU Leuven (Leuven, Belgium), 2018. Proc. of the 28th IEEE International Workshop on Machine Learning for Signal Processing (MLSP 2018), September, Aalborg, Denmark) [2] M. Boussé, G. Goovaerts, et al., “Irregular heartbeat classification using Kronecker Product Equations,” in Proc. of the 39th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC 2017, Jeju Island, South Korea), 2017, pp. 438–441.