Automated Video Stream Segmentation Using Entropy Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video recording devices and video-on-demand systems face challenges in accurately identifying and segmenting commercial and programming boundaries within video streams, leading to inefficient recording and playback, as they often rely on manual judgment and may miss or record unnecessary content due to fixed time frames.
Innovation Solution
Automated segmentation of linear content streams using image analysis measurements like entropy and mutual information to identify cuts, fades, and programming boundaries, combined with audio duplicate detection and logo recognition to classify content types and adjust recording times accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual judgment is used to identify commercial boundaries in video streams, then classification accuracy may be improved, but labor intensity and time consumption increase significantly
Solution Approach 1:
The patent replaces manual mechanical judgment with automated image processing and analysis systems. Specifically, it uses entropy calculation, mutual information measurement, and template matching algorithms to automatically detect commercial boundaries, eliminating the need for manual viewing and classification while maintaining high accuracy through computational methods.
Solution Approach 2:
The system enables self-service by allowing the video processing system to automatically identify and classify commercial segments without human intervention. The automated algorithms analyze video frames, calculate entropy values, and determine commercial boundaries independently, making the process autonomous and eliminating labor-intensive manual operations.
2Device complexity
If fixed time frames are used for video recording, then device simplicity is maintained, but recording efficiency decreases due to wasted storage space or incomplete program capture
Solution Approach 1:
The patent introduces dynamic adjustment capabilities to the recording system. Instead of fixed time frames, the system dynamically determines recording boundaries by analyzing video content characteristics, entropy changes, and commercial segment identification. This allows the recording duration and boundaries to adapt automatically to the actual program length and structure, optimizing storage utilization and ensuring complete program capture.
Solution Approach 2:
The system performs preliminary analysis of the video stream to identify program boundaries and commercial segments before finalizing recording parameters. By pre-processing the video content to detect entropy changes and mutual information patterns, the system can determine optimal recording start and end times in advance, ensuring efficient storage allocation and complete program recording without fixed time constraints.
3Ease of operation
If video streams are segmented into commercial-free content items, then user experience is improved, but processing complexity and computational requirements increase
Solution Approach 1:
The patent applies segmentation by dividing the video stream into distinct segments based on detected commercial boundaries. The system identifies and separates commercial segments from program content using entropy analysis and mutual information measurements, creating discrete, manageable segments that can be independently processed, stored, and delivered to users without commercials.
Solution Approach 2:
The system utilizes parameter changes in video content characteristics to identify commercial boundaries. By monitoring changes in entropy values, mutual information, and frame differential patterns, the system detects transitions between program and commercial content. These parameter-based detection methods provide automated, accurate segmentation without requiring complex manual intervention.
4Extent of automation
If automated segmentation algorithms are implemented, then labor requirements are reduced, but measurement precision requirements increase to accurately distinguish commercial from program content
Solution Approach 1:
The patent introduces intermediary computational measures such as entropy calculation and mutual information as mediators between the video content and the segmentation decision. These intermediary metrics serve as objective, quantifiable indicators that bridge the gap between raw video data and commercial boundary identification, enabling automated algorithms to accurately distinguish commercial from program content through mathematical measurements rather than subjective judgment.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
[ EP 14 163 781 ] Video content streams (210) may be segmented to provide extraction of content items (205a-205d) automatically devoid of commercials or other unrelated content. These content items may then be stored in a database and made accessible to subscribers through, for example, a video-on-demand VOD service. The method comprises: determining (315, 320, 325), by a processing device a plurality of segments (203a-203g) in content (210) based on at least one detected transition in the content; determining, for instance from an electronic program guide EPG, an expected content type of a content item, the content type comprising one of a first content type and a second content type (for instance "news" & "sporting event"); selecting a segment classification algorithm based on the expected content type, wherein the first content type corresponds to a first segment classification (e.g. advertisements) and the second content type corresponds to a second segment classification (e.g. non-advertisements) different from the first segment classification; and categorizing the plurality of segments using the selected segment classification algorithm.