Adaptive habituation detection to build human computer interactive systems using a real-time cross-modal computation

We propose a new habituation detection system using a cross-modal computation. The cross-modal sensory data comprised of eye-movement and skin potential level (SPL) for our habituation detection system has the substantial temporal/spatial nonstationarity. Therefore, it was difficult for conventional classification methods to detect the boundary of the habituation state from the sensory data. Hence, we introduced an Allen-Cahn type partial differential equation (PDE) method to deal with the uncertainty, and developed a new real-time habituation detection system. The result demonstrates that our proposed method performs better classification of the nonstationary data than conventional methods even with a small amount of data.