Neural Network Speech Recognition Methods for Public Speaking Preparing

The purpose of this article is to present an assistant for preparing for public speaking based on speech recognition methods using neural networks. The development of the application is justified by the lack of accessible and effective simulators for speakers. The problem was identified based on surveys of students who have to make presentations of their ideas quite often. The training system uses deep learning methods to recognize and analyze speech, determine its stylistic and emotional coloring, parasite words and filler words, speech speed and provide feedback to the user to improve the sound part of his public speech. The article describes the architecture of the application and provides an implementation scheme. The application has the potential for further development and implementation in the educational environment.