Implementation of a methodology for detecting voice commands using HMM

Abstract

This article presents a solution that allows the identification of voice commands. We built a database with the commands: atrás, adelante, derecha, izquierda, pare and silla. The characterization of the voice commands was conducted with frequency characteristics obtained from the Short Time Fourier Transform, and the recognition of these commands with stochastic GMM and HMM classifiers. The results obtained with this method have a high performance in the order of 98-100% for the recognition of these commands. One possi-ble application of this tool will be the coupling of the recognition system with the engine control conditioned to a wheelchair. This would represent a solution to the problem of mobility in people with motor disabilities, thereby giving greater autonomy achieved in their mobility, improving their quality of life.