Music Emotion Prediction Using Recurrent Neural Networks

TL;DR

The paper tackles music emotion recognition using Russell's Quadrant and compares baseline classifiers with RNN-based networks (RNN, BRNN, LSTM) across small, augmented, and larger datasets. It employs Librosa to extract 14 audio features and evaluates on 4Q and MTG-Jamendo-derived data, revealing that simple baselines can outperform neural nets on small data, while data augmentation and larger datasets enable neural networks, particularly BRNNs and LSTMs, to excel. The study highlights the importance of dataset size for model choice and demonstrates augmentation as a practical route to improve performance. Overall, the work provides insights into temporal modeling for music emotion classification and guides future work on scalable architectures and richer datasets.

Abstract

This study explores the application of recurrent neural networks to recognize emotions conveyed in music, aiming to enhance music recommendation systems and support therapeutic interventions by tailoring music to fit listeners' emotional states. We utilize Russell's Emotion Quadrant to categorize music into four distinct emotional regions and develop models capable of accurately predicting these categories. Our approach involves extracting a comprehensive set of audio features using Librosa and applying various recurrent neural network architectures, including standard RNNs, Bidirectional RNNs, and Long Short-Term Memory (LSTM) networks. Initial experiments are conducted using a dataset of 900 audio clips, labeled according to the emotional quadrants. We compare the performance of our neural network models against a set of baseline classifiers and analyze their effectiveness in capturing the temporal dynamics inherent in musical expression. The results indicate that simpler RNN architectures may perform comparably or even superiorly to more complex models, particularly in smaller datasets. We've also applied the following experiments on larger datasets: one is augmented based on our original dataset, and the other is from other sources. This research not only enhances our understanding of the emotional impact of music but also demonstrates the potential of neural networks in creating more personalized and emotionally resonant music recommendation and therapy systems.

Figures (13)

• Figure 1: Data Visualization for Augmented 4Q Dataset
• Figure 2: Plots of Test Accuracy of Baseline Models
• Figure 3: Unrolled RNNs Models
• Figure 4: Train and Evaluation Accuracy vs. Iterations for RNNs
• Figure 5: Bidirectional RNNs