Application of KNN and SVM classification algorithms in artificial speech recognition by emotion detection in speech signals

Abstract

The creation of modern technologies for detecting emotional changes in individuals in speech signals and providing users with the necessary information for themselves, representing speech signals in digital form, filtering, extracting the necessary features, modelling and analysing them, recognizing a person's voice using intelligent algorithms and software for digital processing, creating voice-controlled devices, examining patients' speech disorders in medicine, classifying speech features, and recognizing a person in speech signals and separating them from artificial speech are urgent issues. To this end, several scientific research works are being carried out aimed at developing and improving methods for separating individual emotions in speech signals and artificial speech using artificial intelligence elements. This article develops artificial speech recognition based on the presence or absence of emotions in speech signals using speech signal features and spectrogram parameters, statistical modelling and classification of speech signals using machine learning algorithms, in particular k-NN and SVM, automatic detection of complex features in speech signals, and analysis and technical solutions for analysing spectrograms of speech data based on deep learning algorithms.