Hardware implementation of a retinal model [recurso electrónico]
García Holguín, Lina María | 2012-11-09
Diseases that degenerate retinal neural cells cause loss of vision because the retina converts light into electrical pulses and performs the first processing stages of the neural visual signals. Thus, the considerable increase over the past decades of these diseases has motivated the development of a retinal prosthesis, capable of restoring functional vision to partial or completely blind patients. Nowadays, significant advances have been accomplished, but a fully functional artificial retina is still a challenge for neuromorphic engineering. In this context, in order to address a research of this nature, this work presents the simulation and hardware implementation of a retinal model, to understand and analyze the signal processing performed by the retinal neural cells and to achieve an optimum methodology for the analysis, development and verification of biological models. Therefore, this thesis starts with the anatomic and physiologic understanding of the retina, and the analysis of several representative retinal models. Afterward it presents the simulation and software verification of two models, and finally the hardware implementation and verification of the retinal model proposed by Z. Pei and Q. Qiao. This model is implemented in the FPGA EP3SL150F1152C2N and its performance is measured with the spike train metrics. Taking into account the implementation results, the hardware retina accomplishes the time and count precision for the spike train output and it can be used as a test vehicle to model the processing of the biological retina.