Scene Change Markers from Crowdsourced Playback Navigation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Streaming video platforms lack automated and programmable scene change markers, making it difficult to efficiently navigate and capture viewer interest, as manual input is impractical for large content catalogs.

Innovation Solution

A system that uses learning systems to automatically generate scene change markers through crowdsourced user interaction data, leveraging machine learning and AI to identify patterns and designate frames of interest as scene change markers, and enhances the playback bar with a sticky navigation feature that snaps to relevant markers based on user interaction patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual input is used for scene change markers, then accuracy of scene change markers is improved, but productivity and scalability worsen due to inability to process large content catalogs

Engineering Contradiction:
Improveaccuracy of scene change markersVSAvoidprocessing speed for large content catalogs
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The system automatically generates scene change markers by analyzing user interaction data from multiple viewers. The learning system processes crowdsourced data to identify frames of interest, eliminating the need for manual input while maintaining accuracy through collective user behavior patterns.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

Manual mechanical input for scene change markers is replaced with an automated learning system that uses machine learning algorithms to analyze user interactions and programmatically generate markers based on aggregated viewer behavior data.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If simple automation is used for scene change markers, then productivity is improved, but measurement precision worsens due to inability to capture viewer interest

Engineering Contradiction:
Improveautomation capability for scene change markersVSAvoidaccuracy in capturing viewer interest
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The learning system continuously collects user interaction data (rewind, fast-forward, pause actions) and uses this feedback to refine and update scene change markers. The system learns from aggregated viewer behavior patterns to accurately identify frames of interest, improving precision while maintaining automation.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

User interaction data serves as an intermediary that bridges simple automation and accurate viewer interest capture. The learning system processes this intermediate data to transform automated marker generation into precision-based scene change identification.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Productivity

If automated learning system is used, then productivity and accuracy are improved, but device complexity increases

Engineering Contradiction:
Improveprocessing capability for scene change markersVSAvoidsystem complexity for implementing learning algorithms
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The learning system is designed to process multiple types of user interactions (rewind, fast-forward, pause) and generate scene change markers across diverse content types. This multi-functionality justifies the complexity by providing a universal solution that handles various scenarios with a single system.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4395336B1Automatically and programmatically generating scene change markers
Publication Date: 2026.02.25 ROKU INC
  • EP4395336B1 patent drawingFigure 1
  • EP4395336B1 patent drawingFigure 2A
  • EP4395336B1 patent drawingFigure 2B

AI summary

Automatically and programmatically generating markers for streaming media in a crowdsourced manner is described herein. The method operates in a media device that is communicatively coupled to a display device. The display device includes a graphical user interface (GUI) having a playback bar. The playback bar includes a current playback time (CTI) and a plurality of markers. The method operates by detecting user interactions as an user navigates through a movie or TV show (or other content) being presented on the display device. Point values are assigned to the user interactions. The detected user interactions and assigned point values represent training data. The training data is provided to a crowdsource server. The crowdsource server is configured to determine correlations and patterns in the training data to automatically and programmatically define crowdsourced markers for the movie/TV show.