A LOW COMPLEXITY METHOD FOR ANALYZING ACOUSTIC ECHO SIGNALS IN THE FRAMEWORK OF A CELLULAR AUTOMATA VIRTUAL ENVIRONMENT

This paper evaluates the accuracy of ultrasound-based object shape classification using a new low complexity method (Binary Transitions) in the context of our recent work on analyzing acoustic signals in virtual environments modeled with cellular automata, which is also reviewed in this paper. This low complexity feature extraction method is compared with other natural computing methods previously designed to optimize solutions for autonomous agents (e.g. robots) localization and orientation. The sound propagation virtual simulator (CANAVI), which was written in both Java and Python, is capable of emulating sound propagation in a controlled 2D environment using Cellular Automata. A review of the various CANAVI implementations and their computational performance is given in this paper. The classifier used in this paper is Fast Support Vector Classifier (FSVC), previously proved to be efficient in isolated speech recognition problems. The results in applying this low complexity transform, based on counting binary transitions, are encouraging in this particular context, showing very good accuracies at a low implementation complexity.