Video Conference Endpoint Face Motion Detection Reframing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Video conference systems often incorrectly assume a participant has left the room when they simply look away or move temporarily, leading to unnecessary camera reframing and disrupting the video experience.

Innovation Solution

Implementing independent face and motion detection techniques, where the system starts a timeout period when a face is no longer detected, and only reframes if no motion coinciding with the face position is detected before the timeout expires, decoupling face detection from motion detection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the system reframes the camera whenever a face is no longer detected, then the camera framing remains updated with current participant positions, but unnecessary reframing occurs when participants temporarily look away or move, disrupting the video experience

Engineering Contradiction:
Improveface detection accuracyVSAvoidparticipant presence detection reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent merges face detection results with motion detection results to make reframing decisions. Instead of relying solely on face detection, the system combines multiple detection mechanisms to distinguish between temporary movements (looking away) and permanent departures, thereby improving the reliability of participant presence detection while maintaining accurate framing.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system uses motion detection as a feedback mechanism to validate or reject face detection losses. When a face is no longer detected, motion detection provides feedback about whether the participant actually moved or left the room, allowing the system to make more reliable reframing decisions and avoid unnecessary disruptions.

Inventive Principle:
Principle #23Feedback

2Device complexity

If the system uses only face detection to determine participant presence, then the implementation is simple, but the system incorrectly assumes participants have left when they temporarily move or look away

Engineering Contradiction:
Improvedetection system complexityVSAvoidparticipant presence detection reliability
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent combines face detection and motion detection into a unified participant presence determination system. This merging of detection methods improves reliability by cross-validating results from multiple sources, while the integration is designed to manage complexity through shared processing infrastructure and coordinated decision-making logic.

Inventive Principle:
Principle #5Merging (Combining)

3Reliability

If the system waits to confirm participant departure before reframing, then unnecessary reframing is avoided, but the response time to actual departures is delayed

Engineering Contradiction:
Improvereframing decision accuracyVSAvoidresponse time to participant departure
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary motion detection when a face is no longer detected, before committing to a reframing decision. This preliminary action allows the system to quickly rule out temporary movements (where motion is detected but the participant remains in frame) and only proceed with reframing when motion confirms actual departure, thus maintaining both accuracy and reasonable response time.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3130138B1Use of face and motion detection for best view framing in video conference endpoint
Publication Date: 2021.06.02 CISCO TECHNOLOGY INC
  • EP3130138B1 patent drawingFigure 1
  • EP3130138B1 patent drawingFigure 2
  • EP3130138B1 patent drawingFigure 3

AI summary

A video conference endpoint detects faces at associated face positions in video frames capturing a scene. The endpoint frames the video frames to a view of the scene encompassing all of the detected faces. The endpoint detects that a previously detected face is no longer detected. In response, a timeout period is started and independently of detecting faces, motion is detected across the view. It is determined if any detected motion (i) coincides with the face position of the previously detected face that is no longer detected, and (ii) occurs before the timeout period expires. If conditions (i) and (ii) are met, the endpoint restarts the timeout period and repeats the independently detecting motion and the determining. Otherwise, the endpoint reframes the view to encompass the remaining detected faces.