A low complexity method based on reaction-diffusion transform for ultrasound echo-based shape object classification

Abstract

This paper presents improvements in terms of accuracy for shape object classification using a new low complexity method compared to previous implementation [1]. The method is using echoes generated by a JAVA platform capable of emulate sound propagation in a controlled 2D virtual environment [2][3]. Echoes originate from the ultrasonic waves generated inside a virtual environment which contains geometrical shape objects. The low complexity method is called RDT (Reaction Diffusion Transform) previously proved efficient in isolated speech recognition problems [4]. The classifier employed in this paper is also a low-complexity one (Fast Support Vector Classifier) previously developed by us in C++ and interfaced with Octave. Results are quite encouraging with 100% accuracy in discriminating circular versus square objects independent on their distance from the ultrasound speaker. For a set of 4 different shapes, the average accuracy is better than 84%.