Master Tesia

Speech emotion recognition (SER) is one of the most popular research directions nowadays, however, most of the solutions suggested by the researchers are tested on different databases and different features, which makes it impossible to compare their results. In this work, we aim to compare three of the most commonly used speech emotion classifiers on the database of spontaneous emotional speech to find the most successful one. Three feature sets have also been used in attempt to find the best speech signal representation. After a number of experiments, SVM has turned out the best of the three solutions on the classification of two and three classes of emotions. Keywords: emotion, speech emotion recognition, SER systems, SVM, DNN, LSTM