Detecting emotions from human speech: role of gender information

For a human being, emotion is an expressive gesture through which we communicate and understand each other. Emotions constitute a major part in our day-to-day communication and other activities. In this work, we implement speech-based emotion recognition (SER) systems to study the role of gender information and augmentation on system performance. Experiments are conducted on a dataset prepared by combination of four publicly available datasets, Toronto Emotional Speech Set (TESS), Surrey Audio-Visual Expressed Emotion (SAVEE), Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS), and Crowd-Sourced Emotional Multimodal Actors Dataset (CREMA-D). Results indicate the efficacy of LSTM based approach specifically on female gender on augmented dataset.