Evaluation of the Effect of Frame Size on Speech Emotion Recognition

Speech emotion recognition aims to determine the emotional state in people by processing the speech with digital signal processing methods. Speech emotion recognition is influenced by many factors such as the language, the demographic characteristics of the participants, and the signal processing methods used. By pre-processing the speech signals before emotion recognition, the signal is improved and the recognition performance is being increased. The framing involved in preprocessing methods is the division of the speech signal into small pieces. In this study, we investigated the effect of the frame size used in framing on the emotion recognition. In addition, the most appropriate frame size for the most commonly used datasets in the literature has been determined. According to the results obtained, the frame size to be used differs according to the dataset used. Furthermore, even the 1ms difference in frame size causes a change in the rate of emotion recognition.