Music Genre and Emotion Recognition Using Both Audio and Textual Features Analysis

Audio content analysis is about summarizing features of audio and classify them. Structural analysis is about high-level things like predicting tags (Genres), recommendation systems, search for cover ID. Most of the early-stage automatic Music Emotion Recognition (MER) systems were based on audio content analysis. Later on, researchers started combining audio and lyrics, leading to bi-modal MER systems with improved accuracy. In this paper, we proposed a novel method which combines the both audio and lyrics (textual) features with LDA topic model for MER followed by a support vector machine. Experimental results showed that the proposed method is more accurate than the baselines.

Keywords:

Music Genre Classification , Music Emotion Recognition , Audio Features , Lyrics Analysis , LDA Topic Model