Audio Analysis for Chapter Detection in Multimedia Streams
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional commercial detection techniques in multimedia streams are ineffective due to anti-skip measures implemented by content providers, making it difficult to accurately identify and skip advertisements in multimedia programs.
Innovation Solution
A multimedia processing system that employs audio analysis to identify chapter breaks by detecting changes in high frequency edges, total energy in the central frequency band, and bitrate patterns, allowing for the characterization of intervals as potential or actual chapter breaks, and implementing a commercial skip function.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional commercial detection techniques (video scene changes or black frames) are used, then the system structure is simple, but the detection accuracy deteriorates due to anti-skip measures
Solution Approach 1:
The patent replaces video-based detection mechanisms with audio-based detection. Specifically, it uses audio spectrum analysis, energy detection, and pattern recognition in the audio domain to identify commercials, substituting the conventional video scene change detection and black frame detection methods that are easily defeated by anti-skip measures.
Solution Approach 2:
The patent changes the detection parameters from visual parameters (video frames, scene changes) to acoustic parameters (audio spectrum, energy levels, frequency patterns). It analyzes audio characteristics such as spectrum changes, energy thresholds, and repeating patterns to detect commercial boundaries, fundamentally changing the parameter space in which detection occurs.
2Measurement precision
If audio analysis with multiple parameters (high frequency edges, total energy, bitrate patterns) is implemented, then the commercial detection accuracy is improved, but the processing complexity increases
Solution Approach 1:
The patent segments the audio analysis into distinct functional modules: high frequency edge detection, total energy analysis in central frequency band, bitrate pattern recognition, and interval characterization. Each module processes a specific aspect of the audio signal, and their results are combined to make detection decisions, dividing the complex task into manageable segments.
Solution Approach 2:
The patent adds multiple dimensions to the audio analysis by simultaneously examining frequency domain characteristics (high frequency edges), energy domain characteristics (total energy in central band), and temporal patterns (repeating values, bitrate changes). This multi-dimensional approach improves detection accuracy by capturing different aspects of commercial vs. program content.
Data Source
AI summary
A multimedia processing system identifies chapter breaks in a program represented by multimedia data through an analysis of audio content of a portion of the multimedia data so as to identify an interval that potentially represents a chapter break. This audio analysis can include an analysis to identify changes in high frequency edges in the audio content, an analysis to identify changes in the total energy in a central frequency band of the audio content, an analysis to identify patterns of sequentially repeating values in the audio content, an analysis to identify changes in bitrate, or some combination thereof. One or more variances in the audio information at (e.g., before) the start of the interval and the audio information at (e.g., after) the end of the interval then may be used to identify or otherwise characterize the interval as representing an actual chapter break. Further, a chapter represented by two consecutive chapter breaks can be identified as an advertisement based on the duration between the two consecutive chapter breaks and thus the multimedia processing device can implement a “commercial skip function” by omitting playback of the portion of the multimedia data representing the chapter responsive to the chapter being identified as an advertisement.