Compressed Domain Music Mood Classification via MDCT Extraction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional music mood classification methods are slow due to the need for decoding compressed music files to extract features like timbre and tempo, leading to inefficient processing speeds and many classification errors, especially when mood classes are defined regardless of genres, resulting in unsophisticated user experiences.

Innovation Solution

Extracting Modified Discrete Cosine Transformation (MDCT)-based timbre and tempo features directly from the compressed domain of music files, followed by genre-based classification and reclassification of uncertain categories to improve accuracy and reliability, with features like spectral centroid and modulation spectrum being used for mood classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If music files are decoded to extract features from decompressed domain, then feature extraction accuracy is improved, but processing speed deteriorates

Engineering Contradiction:

Improvefeature extraction accuracyVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSSpeed

Solution Approach 1:

The patent extracts only the necessary MDCT coefficients directly from the compressed domain without full decoding, taking out only the essential spectral information needed for mood classification while leaving the rest of the decoding process unnecessary, thus achieving both accuracy and speed

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary extraction of MDCT coefficients from the compressed domain before any decoding occurs, preparing the essential features in advance for mood classification, which eliminates the need for time-consuming full decoding while maintaining feature quality

Inventive Principle:

Principle #10Preliminary action

2Device complexity

If mood classes are defined regardless of genres, then classification simplicity is improved, but classification accuracy deteriorates

Engineering Contradiction:

Improveclassification system simplicityVSAvoidmood classification accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the classification process into two stages: first classifying by genre, then by mood within each genre category. This segmentation allows the system to handle complexity in an organized manner, improving accuracy without overwhelming the system

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies different classification criteria and features locally to each genre category. For example, different mood classes and feature weights are used for different genres, allowing each category to be classified with the most appropriate criteria for that specific genre, thereby improving overall accuracy

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS7582823B2Method and apparatus for classifying mood of music at high speed

Publication Date: 2009.09.01 SAMSUNG ELECTRONICS CO LTD

AI summary

A method and apparatus for classifying mood of music at high speed. The method includes: extracting a Modified Discrete Cosine Transformation-based timbre feature from a compressed domain of a music file; extracting a Modified Discrete Cosine Transformation-based tempo feature from the compressed domain of the music file; and classifying the mood of the music file based on the extracted timbre feature and the extracted tempo feature.