Finger position detection using multitask Gaussian process regression on noncontact control panels

In recent years, the development of transparent antennas for fifth-generation mobile communication systems has progressed, in which transparent conductive films are mounted on glass. To investigate transparent antenna applications apart from telecommunications, we have studied algorithms to realize noncontact control panels using transparent antennas. In this study, we report the development of an algorithm that can estimate the finger position three-dimensionally using multitask Gaussian process regression, which is a machine learning method based on multidimensional Gaussian inference. This algorithm takes advantage of the antenna frequency response to a finger approaching transparent antennas arranged in a two-dimensional manner within a noncontact control panel. The effectiveness of the proposed method was verified by analyzing simulation and experimental data acquired using a dedicated testbed developed for this study.