A Real-time Emotion Recognition from Speech using Gradient Boosting

Emotion recognition from speech is one of the research fields for emotional human-computer interaction. In this contribution, a real-time emotion recognition system is presented which recognizes emotions from live recorded speech by analyzing tonal properties. 34 audio features are extracted including MFCCs, energy, spectral entropy etc. Basically this system classifies emotions using models trained by Gradient Boosting. Other two classifiers such as Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) are also applied to observe the accuracies of them on test audio files. Two databases are employed for training the system like RA VDESS and SA VEE database. This system examines four basic emotions - anger, happiness, sadness and neutral.