Figure 5
From: Multi-modal self-adaptation during object recognition in an artificial cognitive system
Blurred vision detection through Molyneux mechanism and filtering. By applying the Molyneux mechanism to every visual-haptic pair, the ACS can determine if the two channels are matching or not, hence it can detect failures in one channel. Although there is a clear trend towards the two channels matching when both are working properly, short mismatches can appear (blue plot). In order to obtain an accurate decision whether the channels are matching or not, a low pass filter is applied to attenuate these mismatches (orange plot). Finally, as a way to offer a binary response that decides if the vision is blurred or not, the filter output goes through a hysteresis cycle (green plot). With this entire process the detection of the blurred vision is not immediate and has an average delay of 2.13 s (85.33 samples). Note that there is no mismatch in lat00 and lon05, that is, for these two objects the blur filter does not affect the recognition system.
