10/04/2014
José Manuel Ferrández Vicente
Diseño Electrónico y Técnicas de Tratamiento de Señal
Electrónica, Tecnología de Computadores y Proyectos
Escuela Técnica Superior de Ingeniería de Telecomunicación
Universidad Politécnica de Cartagena
Cartagena
España
This thesis deals with two different fields, inherently related to each other in this case: neuroscience and computation. The overall objetive of this thesis is the development of a biological neuroprocessor with cultured biological neural networks using microelectrode arrays as platform. In this way, a real-time close-loop experimentation system with neural cultures has been proposed, wich provides a complete solution for filtering, visualization, processing and stimulation of electrophysiological response from neural population and comunication with a robotic system. Centre of area algorithm for robotic guidance has been adapted to the functional response of neural populations, identifying those electrodes from the array whose neurons increase the most its firing rate, as target for robotic guidance.
Statistic calibration and normalization techniques of neural population registers have been developed for supressing the intrinsic variability of those populations and the different-non-homogeneity characteristics, both in culture density and electrical properties of the electrodes. Finally, a natural learning paradigm has been proposed, Hebbian learning, to conform functional connections between previously not connected electrodes.
