2D DCT Audio Classification for Techno Genre Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio classification methods face challenges in accurately identifying musical genres, particularly techno music, due to their dependence on general classification methods and features that may not effectively capture the unique characteristics of non-melodic audio signals, leading to difficulties in audio matching systems.
Innovation Solution
The proposed solution involves generating a spectrogram and two-dimensional discrete cosine transforms (DCTs) for an audio sample, normalizing the max values based on total frequencies, and aggregating scores across consecutive windows to identify the presence of techno music by setting a genre threshold, thereby improving the precision of audio matching systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If general classification methods are used for audio genre detection, then the system can handle diverse audio types, but it fails to accurately identify specific genres like techno music
Solution Approach 1:
The patent segments the audio analysis process into distinct components: generating a spectrogram, applying 2D DCT to the spectrogram, extracting energy values from specific columns, and aggregating scores. This segmentation allows the system to handle diverse audio types through the general spectrogram generation while applying specialized 2D DCT-based energy analysis specifically for techno music detection, thereby resolving the contradiction between versatility and precision.
Solution Approach 2:
The patent applies local quality by focusing analysis on specific regions of the 2D DCT output - particularly columns 2-10 which correspond to specific frequency ranges characteristic of techno music. Instead of uniformly analyzing the entire spectrogram for all audio types, the system selectively examines local regions that are most indicative of techno genre characteristics, improving precision for techno detection while maintaining general applicability.
2Reliability
If feature extraction methods are designed for melodic music, then they work well for melodic genres, but they fail to capture characteristics of non-melodic techno music
Solution Approach 1:
The patent changes the analysis parameters by transforming the spectrogram through 2D DCT and then analyzing energy distribution in specific column regions rather than using traditional melodic feature extraction methods. This parameter transformation allows the system to detect patterns characteristic of non-melodic techno music (rhythmic patterns, beat structures) that are invisible to conventional melodic analysis, thereby extending versatility to non-melodic genres while maintaining reliability through the mathematically rigorous DCT transformation.
3Productivity
If traditional audio matching is used, then it can identify audio samples based on general features, but it produces false matches for techno music with non-melodic patterns
Solution Approach 1:
The patent introduces 2D DCT transformation as an intermediary step between spectrogram generation and feature extraction. This intermediary transformation converts the spectrogram into a different representation where techno music characteristics become more distinguishable. The 2D DCT acts as a mediator that preserves the computational efficiency of traditional matching while improving identification accuracy by revealing patterns specific to techno genre in the transformed domain.
Data Source
AI summary
Systems and methods are provided herein relating to audio classification. Genres of music can be identified by detecting unique spectral features inherent to those genres. One example genre detected is techno music. Two dimensional discrete cosine transforms can be generated for consecutive windows of the spectrogram or chromagram. A max value of the energy of portions of the two dimensional discrete cosine transforms can be determined. The max value can be normalized and aggregated with max values related to neighboring windows. If the aggregate scores meet a genre threshold, the audio sample, or portions thereof, can be associated with a genre of music.