Audience Measurement Using Trick-Mode Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current audience measurement systems fail to accurately assess viewer engagement with video content using trick-mode data, as they lack efficient methods to analyze and interpret trick-mode events for determining engagement levels in video sequences.
Innovation Solution
A method involving the collection and aggregation of trick-mode start and end timestamps from end-user devices, merging adjacent trick-mode events, and analyzing clusters to identify engagement levels in video content, where fast-forwarded content is considered low engagement and rewound content is high engagement, allowing for the sharing of engagement data with content providers and recommendation engines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If trick-mode events are collected and analyzed to determine viewer engagement, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts only the essential trick-mode event data (start timestamps, end timestamps, and event types) from end-user devices and sends this minimal dataset to the audience measurement system for analysis. This extraction approach improves measurement precision by focusing on relevant engagement indicators while avoiding the complexity of processing complete playback sequences or detailed user interactions.
Solution Approach 2:
The audience measurement system segments trick-mode data into discrete events with specific attributes (start time, end time, event type). By dividing continuous playback data into segmented trick-mode events, the system achieves precise engagement measurement through structured analysis while maintaining manageable data complexity through standardized event formatting.
2Loss of information
If trick-mode timestamps are aggregated and clustered to identify engagement levels, then information quality is improved, but processing time increases
Solution Approach 1:
End-user devices perform preliminary actions by collecting and timestamping trick-mode events locally before transmission. The system pre-processes raw playback data into structured trick-mode events with embedded timestamps, so that when data reaches the audience measurement system, aggregation and clustering operations can proceed efficiently without requiring complex real-time processing of raw playback streams.
Solution Approach 2:
The patent applies partial action by selecting only specific trick-mode events for detailed analysis rather than processing every playback event. The system identifies and analyzes clusters of trick-mode events that are most indicative of engagement patterns, performing sufficient analysis to achieve high information quality while avoiding excessive processing of all possible data points.
3Measurement precision
If fast-forward and rewind behavior is analyzed to determine engagement, then measurement accuracy is improved, but system complexity increases
Solution Approach 1:
The patent applies inversion by analyzing trick-mode events in reverse chronological order or by examining end timestamps first to identify engagement patterns. The system inverts the traditional approach of analyzing playback from start to finish, instead using end timestamps and reverse sequencing to quickly identify high-engagement content sections, thereby improving measurement accuracy while reducing analysis complexity.
Data Source
AI summary
In one embodiment, a method includes receiving trick-mode start timestamps of trick-mode events performed on a video sequence in end-user devices such that each trick-mode start timestamp is associated with a trick-mode event performed on a video sequence in one end-user device, receiving trick-mode end timestamps of the trick-mode events, aggregating the trick-mode start timestamps according to a time value of each of the trick-mode start timestamps, aggregating the trick-mode end timestamps according to a time value of each of the trick-mode end timestamps, identifying a plurality of start clusters from the aggregation of the trick-mode start time stamps, identifying a plurality of end clusters from the aggregation of the trick-mode end time stamps, analyzing the plurality of start clusters and the plurality of end clusters, and identifying a level of engagement of a section of the video sequence based on the analyzing.