Astigmatism-based displacement sensor powered by deep neural networks

Abstract

As high-sensitivity sensors with astigmatism can be implemented with a compact volume, they are widely used as detection sensors that measure qualitative changes on nano- and micrometer scales. However, owing to problems such as slow measurement speed and a decrease in accuracy due to nonlinearity, its utilization as a displacement sensor is limited. Therefore, an approach to improve the speed and accuracy of measurements is proposed in this paper. We introduce a displacement sensor using a digital versatile disc (DVD) pickup head based on the principle of astigmatism. A method of correcting the nonlinear characteristics of the astigmatism-based DVD pickup head at high speed using a deep neural network is proposed. The proposed system was implemented on a microcontroller unit that can be fabricated in the form of a modular sensor. In the experiments, the proposed method showed better performance in terms of accuracy and speed than the conventional ninth-order polynomial regression.