Multi-Task Feature Selection for Speech Emotion Recognition: Common Speaker-Independent Features Among Emotions

Feature selection is one of the most important steps in designing the speech emotion recognition systems. Since there is uncertainty as to which speech feature is related to which emotion, many features must be taken into account, and for this purpose, identifying the most discriminative features is necessary. In the interest of selecting the appropriate emotion-related speech features, in the current work, we focus on a multi-task approach. For this reason, we consider each speaker as a task, and propose a multi-task objective function in order to select the features. As a result, the proposed method chooses one set of speaker-independent features, of which the selected features are discriminative in all the emotion classes. Correspondingly, the multi-class classifiers are utilized directly or the binary classifications simply perform the multi-class classifications. In addition, we employ two well-known datasets, Berlin and Enterface. The experiments are also applied on the openSmile toolkit in order to extract more than 6500 features. After the feature selection phase, the results obtained illustrate that the proposed method selects the features that are common in different runs. Also the runtime of the proposed method is the lowest in comparison to the other methods. Finally, seven classifiers are employed; the best achieved performance is 73.76% for the Berlin dataset and 72.17% for the Enterface dataset in the face of a new speaker. These experimental results then show that the proposed method is superior to the existing state-of-the-art methods.