Mobile robot controlled by voice

Abstract

All of the state-of-the-art systems are statistical data-driven systems that rely on training to model our ignorance about speech which result in enormous memory and processor consumption. Those solutions can not be easily integrated in embedded systems like e.g. as interface for human-robotic interaction. In this paper, an isolated word recognition system has been developed for communication with mobile robot. Presented algorithm is based on phoneme recognition by identifying characteristics of each phoneme which means that it's not based on statistical ASR (not based on HMM-Hidden Markov Model or ANN-Artificial Neural Network). Final goal of this project is developing a speaker and language independent phoneme recognition platform for use in embedded systems with low power and memory consumption.