Resumen

FERRIN-BOLANOS, Carlos; LOAIZA-CORREA, Humberto; PIERRE-DIAZ, Jean  y  VELEZ-ANGEL, Paulo. Assessing the Contribution of Covariance Information to the Electroencephalographic Signals of Brain-Computer Interfaces for Spinal Cord Injury Patients. TecnoL. [online]. 2019, vol.22, n.46, pp.158-176. ISSN 0123-7799.  https://doi.org/10.22430/22565337.1392.

Non-invasive EEG-based motor imagery brain-computer interfaces (miBCIs) promise to effectively restore the motor control of motor-impaired patients with conditions that include Spinal Cord Injury (SCI). Nonetheless, miBCIs should be further researched for this type of patients using low-cost EEG acquisition devices, such as the Emotiv EPOC, for home rehabilitation purposes. In this work, we describe in detail and compare ten miBCI architectures based on covariance information from EEG epochs. The latter were acquired with Emotiv EPOC from three control subjects and two SCI patients in order to decode the close and open hand intentions. Four out of the ten miBCIs use covariance information to create spatial filters; the rest employ covariance as a direct representation of the EEG signals, thus allowing the direct manipulation by Riemannian geometry. We found that, although all the interfaces present an overall accuracy above chance level, the miBCIs that use covariance as a direct representation of the EEG signals together with linear classifiers outperform miBCIs that use covariance for spatial filtering, both in control subjects and SCI. These results suggest the high potential of Riemannian geometry-based miBCIs for the rehabilitation of SCI patients with low-cost EEG acquisition devices.

Palabras clave : Brain-computer interfaces; Motor imagery; Information geometry; Spatial filters; Spinal cord injury.

        · resumen en Español     · texto en Español     · Español ( pdf )