Ultrasonic pattern recognition based on ANN and PCA techniques

The reflector classification has been used in ultrasonic sensorial systems to develop tasks, as environment mapping or navigation items. This work proposes a new ultrasound-based sensorial structure, where times-of-flight are used for classification purposes with an artificial neural network (ANN). PCA technique is used before to reduce the data amount managed by the neural network. The system allows the discrimination among planes, edges and corners. The recognition is achieved with an error of 5%.