Abstract
Ultrasonic sensors are often utilized to perceive the outside world. In this paper, we propose the construction of an object identification system that utilizes the ultrasonic waves reflected from a measured object. This identification system also detects the position and posture of an object. The system consists of an ultrasonic sensor array and a neural network. The ultrasonic sensor array is made up of plural ultrasonic transmitters and receivers that are arranged on the same plane. The receiver outputs the peak values of ultrasonic sound pressure waves. Their values are input into the neural network, which identifies the shape of the object and detects its position and posture. We have designed and built a prototype system and carried out numerous experiments using objects with primitive shapes or transparent bottles (PET bottles). The experimental results showed almost perfect shape identification and a measurement accuracy of up to one-half of one wavelength of the ultrasonic wave.
