Since 2004, I have been working toward developing brain-computer interfaces for the compensation of severe motor disabilities. Before that, during almost 10 years, I have developed techniques for processing and analyzing image sequences and pairs of stereoscopic images.
Brain-machine interfaces
My first research axis, followed in collaboration with Marie-Hélène Bekaert and Claudine Lecocq, concerns ergonomy and adaptation of these HMI to the specificity of patient’s handicap. We are conducting this research with colleagues of Prof. André Thévenon’s team, especially with Dr. Vincent Tiffreau (Physical and Functional Rehabilitation unit, Lille Regional Hospital). This research field offers many opportunities, especially with respect to the development of long term adaptive aids for patients suffering from Duchenne’s muscular dystrophy (DMD).
On the other hand, I work in collaboration with the Department of Neurosurgery of the Laennec Hospital in Nantes, headed by Prof. Jean-Paul Nguyen, on the performance of brain-computer interfaces using signals recorded by epidural electrodes. Two experiments were conducted in 2009 and 2011 on patients undergoing treatment of neuropathic pain by chronic stimulation of the motor cortex with implanted electrodes. The therapeutic goal is to enable a more efficient and faster functional recovery using the robust neuro-feedback elaborated by the BCI.
Image analysis
Early in my career, I have developed algorithms for processing stereo image pairs and/or image sequences, mainly in order to improve computational efficiency. With Madain Perez-Patricio, I evaluated the possibility of implementing dense stereo algorithms in hardware and proposed an algorithm that uses neighborhoods of varying sizes and shapes to estimate the similarity. With Sébastien Lefebvre, we have proposed a method allowing the real-time processing of stereoscopic image pairs, this time on the basis of mono-dimensional windows. I then co-supervised, with Ludovic Macaire, Hachem Halawana’s PhD thesis, who performed stereo matching directly on images formed by color cameras equipped with single CCD sensors (CFA matrix).
I also addressed the problem of matching a parametric 3D model of an object with the images of the latter. In the PhD thesis of Wei Wu, adjusting the parameters of the model allowed us to measure the angular position of an object freely rotating about a fixed point. In the PhD thesis of Yang Rong, we again used a parametric model to locate a vehicle on the road lanes, this time using rear-facing cameras. The accurate modeling of objects with geometric and photometric descriptions, that allows one to take into account reflections, was then the subject of Yara Bachalany’s PhD thesis.
Version française