Implementation of a model for automatic recognition of human emotions from speech

Abstract

Implementation of a model for automatic recognition of human emotions from speech

Baryshev D.A., Zubankov A.S., Rozaliev V.L.

Incoming article date: 01.03.2024

Determining human emotions from speech is a pressing task at the moment, because it can be applied in various industries, such as economics, medicine, marketing, security and education. This work examines the recognition of human emotions specifically from speech, because speech is an informative indicator that is quite difficult to fake. The paper discusses a neural network approach to solving the problem. A recurrent neural network with LSTM memory was implemented, and our own dataset was collected on which the model was trained. The dataset includes the speech of Russian-speaking actors, which will improve the quality of the model for Russian-speaking users.

Keywords: neural network, emotion detection, speech, classification, deep learning, recurrent model, LSTM