We aim to investigate biological function of emotional bodily movements while building a robot in the sense of survival. In this particular paper, we build a novel active vision system which detects weak signals by generating tremor actively. Proposed active vision system consists of a noise generator and a neural system. As a noise generator, we use motors with a decentered weight. And as a neural system in order to detect visual signals, we use ensemble of FitzHugh-Nagumo neurons. We first show that our robotic platform can generate several kinds of tremors; tremor around 40 Hz, 50 Hz, and tremor which does not involve specific peak in its power spectrum. And, in order to evaluate the effect of tremor, we prepare clear-films as visual stimuli which move aperiodically in a dark environment. We show that as a result of sensory noise induced by tremor, and as a result of stochastic resonance, neural system successfully detects movements of clear films.