XGBoost-based Music Emotion Recognition with Emobase Emotional Features

The growing integration of music into daily life, especially for relaxation and stress relief, highlights the need for innovative methods to analyze and categorize emotional responses to music. This study explores a straightforward approach to music emotion recognition (MER) using emotional features extracted by Emobase (openSMILE) and the eXtreme Gradient Boosting (XGBoost) algorithm in Emopia2.2 Dataset(1073 wav files). This method operates within Russell’s Circumplex four-quadrant (4Q) emotion classification framework and has proven effective, with the XGBoost model achieving a classification accuracy of 71.16%. This performance surpasses the 69.2% accuracy of traditional multimodal Multifaceted MER approaches, highlighting the effectiveness of this simplified method in improving emotion classification in music.