Building robot voice control training methodology using artificial neural net

Abstract

The article describes using artificial neural net for speech recognition. This is necessary for the automation of construction work. We selected multi-layer perceptron for separate words recognition tasks. We researcheddifferent forms of hidden layer. And we made a conclusion that for voice commands analysis tasks solution it is expedient to use multi-layer perceptron with linearized functions (the best result was achieved using model with hidden layer with linearized hyperbolic tangent function). We use this experience to train robots in civil engineering.