DESIGNING AN INFORMATION SYSTEM WITH THE POSSIBILITY OF VOICE CONTROL

Abstract

The work is devoted to the creation of an information system for recognizing voice commands based on artificial neural networks. With the development of computer systems, it is becoming more and more obvious that the use of speech recognition systems will be greatly expanded if it becomes possible to use human language when working directly with the computer, and in particular, it becomes possible to control the machine with a normal voice in real time, as well as input and output information in the form of ordinary human language. One of the promising ways of organizing human-machine interaction is the transmission of user instructions to the computer system in the form of language commands. Voice interface is a necessary component when it comes to creating comfortable living conditions for people with disabilities. In the paper, the approaches to the selection of informative features describing the speech signal are defined: the method of linear prediction and spectral analysis, the structure of a neural network with one feedback is considered, and it is established that the learning of the neural network is carried out by successive presentation of the training sample, with simultaneous adjustment weights according to a specific procedure until the tuning error across the set reaches an acceptably low level. The value of the obtained results lies in the improvement of a new method of speech recognition, which is better adapted to the user's speech, which requires a minimum of resources and the creation of an information system with the possibility of voice control using devices based on various operating systems. An informative cross-platform application with a voice interface was designed based on this approach.