We study the nonlinear FitzHugh-Nagumo model witch describes the dynamics of excitable neural element. It is well known that this system exhibits three different possible responses. Indeed, the system can be mono-stable, oscillatory or bistable. In the oscillatory regime, the system periodically responds by generating action potential. By contrast, in the mono-stable state the system response remains constant after a transient. Under certain conditions, the system can undergo a bifurcation between the stable and the oscillatory regime via the so called Andronov-Hopf bifurcation. In this Phd thesis, we consider the FitzHugh-Nagumo model in the stable state, that is set near the Andronov-Hopf bifurcation. Moreover, we take into account the contribution of noise witch can induces two phenomena coherence resonance and stochastic resonance. First, without external driving, we show the effect of coherence resonance since a critical noise level enhances the regularity of the system response. Another numerical investigation reports how noise can allow to detect a subthreshold deterministic signal applied to the system. In this case, an appropriate amount of noise maximizes the signal to noise ratio reveling the stochastic resonance signature. Besides this numerical studies, we have also built a non linear circuit simulating the FitzHugh-Nagumo model under the presence of noise. This circuit has allowed to confirm experimentally the numerical observation of stochastic resonance and coherence resonance. Therefor, this electronic circuit contributes a framework for further experimental investigation in the field of neural sciences to better understand the role of noise in neural encoding.
