Music Emotion Recognition
Preview
Book Description
Providing a complete review of existing work in music emotion developed in psychology and engineering, Music Emotion Recognition explains how to account for the subjective nature of emotion perception in the development of automatic music emotion recognition (MER) systems. Among the first publications dedicated to automatic MER, it begins with a comprehensive introduction to the essential aspects of MER—including background, key techniques, and applications.
This ground-breaking reference examines emotion from a dimensional perspective. It defines emotions in music as points in a 2D plane in terms of two of the most fundamental emotion dimensions according to psychologists—valence and arousal. The authors present a computational framework that generalizes emotion recognition from the categorical domain to real-valued 2D space. They also:
- Introduce novel emotion-based music retrieval and organization methods
- Describe a ranking-base emotion annotation and model training method
- Present methods that integrate information extracted from lyrics, chord sequence, and genre metadata for improved accuracy
- Consider an emotion-based music retrieval system that is particularly useful for mobile devices
The book details techniques for addressing the issues related to: the ambiguity and granularity of emotion description, heavy cognitive load of emotion annotation, subjectivity of emotion perception, and the semantic gap between low-level audio signal and high-level emotion perception. Complete with more than 360 useful references, 12 example MATLAB® codes, and a listing of key abbreviations and acronyms, this cutting-edge guide supplies the technical understanding and tools needed to develop your own automatic MER system based on the automatic recognition model.
Table of Contents
Introduction
Importance of Music Emotion Recognition
Recognizing the Perceived Emotion of Music
Issues of Music Emotion Recognition
Ambiguity and Granularity of Emotion Description
Heavy Cognitive Load of Emotion Annotation
Subjectivity of Emotional Perception
Semantic Gap between Low-Level Audio Signal and High-Level Human Perception
Overview of Emotion Description and Recognition
Emotion Description
Categorical Approach
Dimensional Approach
Music Emotion Variation Detection
Emotion Recognition
Categorical Approach
Dimensional Approach
Music Emotion Variation Detection
Music Features
Energy Features
Rhythm Features
Temporal Features
Spectrum Features
Harmony Features
Dimensional MER by Regression
Adopting the Dimensional Conceptualization of Emotion
VA Prediction
Weighted-Sum of Component Functions
Fuzzy Approach
System Identification Approach (System ID)
The Regression Approach
Regression Theory
Problem Formulation
Regression Algorithms
System Overview
Implementation
Data Collection
Feature Extraction
Subjective Test
Regressor Training
Performance Evaluation
Consistency Evaluation of the Ground Truth
Data Transformation
Feature Selection
Accuracy of Emotion Recognition
Performance Evaluation for Music Emotion Variation Detection
Performance Evaluation for Emotion Classification
Ranking-Based Emotion Annotation and Model Training
Motivation
Ranking-Based Emotion Annotation
Computational Model for Ranking Music by Emotion
Learning-to-Rank
Ranking Algorithms
System Overview
Implementation
Data Collection
Feature Extraction
Performance Evaluation
Cognitive Load of Annotation
Accuracy of Emotion Recognition
Subjective Evaluation of the Prediction Result
Fuzzy Classification of Music Emotion
Motivation
Fuzzy Classification
Fuzzy k-NN Classifier
Fuzzy Nearest-Mean Classifier
System Overview
Implementation
Data Collection
Feature Extraction and Feature Selection
Performance Evaluation
Accuracy of Emotion Classification
Music Emotion Variation Detection
Personalized MER and Groupwise MER
Motivation
Personalized MER
Groupwise MER
Implementation
Data Collection
Personal Information Collection
Feature Extraction
Performance Evaluation
Performance of the General Method
Performance of GWMER
Performance of PMER
Two-Layer Personalization
Problem Formulation
Bag-of-Users Model
Residual Modeling and Two-Layer Personalization Scheme
Performance Evaluation
Probability Music Emotion Distribution Prediction
Motivation
Problem Formulation
The KDE-Based Approach to Music Emotion Distribution Prediction
Ground Truth Collection
Regressor Training
Regressor Fusion
Output of Emotion Distribution
Implementation
Data Collection
Feature Extraction
Performance Evaluation
Comparison of Different Regression Algorithms
Comparison of Different Distribution Modeling Methods
Comparison of Different Feature Representations
Evaluation of Regressor Fusion
Lyrics Analysis and Its Application to MER
Motivation
Lyrics Feature Extraction
Uni-gram
Probabilistic Latent Semantic Analysis (PLSA)
Bi-gram
Multimodal MER System
Performance Evaluation
Comparison of Multimodal Fusion Methods
Evaluation for PLSA Model
Evaluation for Bi-Gram Model
Chord Recognition and Its Application to MER
Chord Recognition
Beat Tracking and PCP Extraction
Hidden Markov Model and N-Gram Model
Chord Decoding
Chord Features
Longest Common Chord Subsequence
Chord Histogram
System Overview
Performance Evaluation
Evaluation of Chord Recognition System
Accuracy of Emotion Classification
Genre Classification and Its Application to MER
Motivation
Two-Layer Music Emotion Classification
Performance Evaluation
Data Collection
Analysis of the Correlation between Genre and Emotion
Evaluation of the Two-Layer Emotion Classification Scheme
Music Retrieval in the Emotion Plane
Emotion-Based Music Retrieval
2D Visualization of Music
Retrieval Methods
Query by Emotion Point (QBEP)
Query by Emotion Trajectory (QBET)
Query by Artist and Emotion (QBAE)
Query by Lyrics and Emotion (QBLE)
Implementation
Future Research Directions
Exploiting Vocal Timbre for MER
Emotion Distribution Prediction Based on Rankings
Personalized Emotion-Based Music Retrieval
Situational Factors of Emotion Perception
Connections between Dimensional and Categorical MER
Music Retrieval and Organization in 3D Emotion Space
Author(s)
Biography
Yi-Hsuan Yang received a Ph.D. in Communication Engineering from National Taiwan University in 2010. His research interests include multimedia information retrieval, music analysis, machine learning, and affective computing. He has published over 30 technical papers in the above areas. Dr. Yang was awarded MediaTek Fellowship in 2009 and Microsoft Research Asia Fellowship in 2008.
Homer H. Chen received a Ph.D. in Electrical and Computer Engineering from the University of Illinois at Urbana- Champaign. Since August 2003, he has been with the College of Electrical Engineering and Computer Science, National Taiwan University, where he is Irving T. Ho Chair Professor. Prior to that, he held various R&D management and engineering positions with US companies over a period of 17 years, including AT&T Bell Labs, Rockwell Science Center, iVast, and Digital Island. He was a US delegate for ISO and ITU standards committees and contributed to the development of many new interactive multimedia technologies that are now part of the MPEG-4 and JPEG-2000 standards. His professional interests lie in the broad area of multimedia signal processing and communications.
Dr. Chen is an Associate Editor of IEEE Transactions on Circuits and Systems for Video Technology. He served as Associate Editor of IEEE Transactions on Image Processing from 1992 to 1994, Guest Editor of IEEE Transactions on Circuits and Systems for Video Technology in 1999, and an Associate Editorial of Pattern Recognition from 1989 to 1999. He is an IEEE Fellow.