Artificial Neural Network based Automatic Speech Recognition Engine for Voice Controlled Micro Air Vehicles

Abstract

Voice Controlled MAV (Micro Air Vehicle) is an attractive alternative to flying the MAVs without a joystick/ mouse clicks. This being a command and control application calls for accurate and fast Speech Recognition. The paper proposes a feed forward neural network based speech recognition for voice controlled MAV application, which achieves better accuracy and faster recognition compared Viterbi algorithm which operates on statistical data. ANN (Artificial Neural Network) could achieve word accuracy of 93% against 85% as achieved by HMM (Hidden Markov Model). ANN achieved about 25% faster recognition compared to HMM.