Using Convolutional Neural Networks for Music Emotion Recognition on Microcontrollers

This paper presents a theoretical overview of music emotion recognition (MER) and its processes. More than 30 studies related to emotion recognition and MER were analyzed. Summaries of information about emotion recognition methods, available audio datasets, and audio parameters used in MER are discussed in this paper. The practical part is focused on a proof of concept for MER using the Convolutional Neural Network model on microcontrollers for realtime music emotion classification in four emotion classes based on a 2D emotion model.