APPLICATION OF NONLINEAR DYNAMICS METHODS FOR JOY EMOTION RECOGNITION IN SPEECH
Read the full article
The article deals with the task of recognizing speech samples, recorded when speakers under test display the emotion of joy, from samples of the same speakers’ speech in a neutral state. Methods of nonlinear dynamics were used to solve the problem. Studies were performed on recordings from the base of Emo-DB (Berlin) and fragments of the Russian bases (Tver). A model housing of emotional speech was formed, consisting of twolevel database (sentences and phonemes), served as the basis for the initial evaluation of developed algorithms performance. Stable signs of nonlinear dynamics were formed - the attractor reconstruction and the recurrence plot. New quantitative signs were proposed for speech patterns classification of the person experiencing the joy emotion, based on estimates of the maximum vectors of the attractor reconstruction for the four quadrants.