Le 8 octobre 2015, le GIPSA-lab accueille Tatiana Aksenova (Clinatec Grenoble) pour un séminaire intitulé « Tensor based decoding of brain neuronal activity for BCI project in Clinatec ». Ce séminaire aura lieu à 15h30 en salle Chartreuse.

Motor Brain Computer Interface (BCI) records brain neuronal signals and converts them to commands for external effectors. Thanks to this, handicapped persons can mentally control a computer cursor, a wheelchair, a robotic arm, an exoskeleton, etc.

BCI project in CLINATEC is based on novel technologies developed at CEA-LETI in the CLINATEC research center. CLINATEC® BCI platform includes a fully-implantable Electrocorticographic (ECoG) recording system, a software environment, signal decoding algorithms, and a 4-limbs exoskeleton EMY (Enhancing MobilitY) dedicated to medical use. ECoG recording systems use electrode grids placed over the cortical surface. ECoG recording device WIMAGINE® has been designed in CLINATEC and manufactured following EC norms for chronic use.

High performance decoding of brain neuronal activity is a key point for successful BCI. Signal processing block of BCI project in CLINATEC is designed to analyze ECoG recordings of tetraplegic subjects during clinical trials. Multi-way (tensor based) analysis is used for limbs trajectories reconstruction. A set of algorithms was invented to respond to BCI challenges of stable, artifact resistant, smooth prediction. Recursive N-way PLS is developed for adaptive BCI system calibration. Penalized NPLS aimed at electrodes and frequency band selection allows group-wise sparsity.

Tensor based approach allows simultaneous data analysis in several domains. Generally, it increases the dimension of the analyzed data (up to 200.000 in CLINATEC® BCI project). Efficient algorithms implementation for real-time brain signal processing ensures high decision rate and minimal system latency. Methods were validated in both invasive (ECoG) preclinical experiments and non-invasive (MEG) BCI study. 

Lieu :

GIPSA-lab salle Chartreuse

11 rue des mathématiques

Domaine universitaire