Texture analysis using a piezoelectric actuator-sensor pair

In this paper, we propose a piezoelectric actuator-sensor pair that can classify several objects. It consists of two polyvinylidene-fluoride films above a polyethylene-terephthalate substrate. Herein, the actuator is connected to an voltage supplier, and the sensor output signal is acquired through a measuring equipment. Specifically, this pair is installed on a robot hand. When the objects are grasped by the robot hand in static state, the actuator oscillates as sinusoidal input voltages with frequency sweep are applied for a few seconds. At the same time, the sensor data is obtained and undergoes preprocessing procedure for learning process. The neural network classifier model is trained by learning process. After conducting the learning process, we test the feasibility of the actuator-sensor pair by demonstrating the real-time recognition system.