A neural network for fast-response ultrasonic distance sensors

Ultrasonic techniques can provide a cost effective solution for distance measurements, but frequently suffer from poor resolution and acoustic noise. In this paper a new approach based on neural networks to process the echo signal is described, it permits the improvement of the resolution of the distance measuring systems by employing ultrasonic techniques and requires a negligible computing time, even though the network, is implemented without special purpose hardware. The results obtained with less-than-one-dollar piezoelectric transducers and a neural network are described. The neural network performance is compared with that of traditional techniques. A sub-millimeter resolution is obtained even in the presence of some acoustic noise, which could prevent other simple detection strategies.