Real-Time Face Recognition via Feature Vector Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Real-time face recognition in live broadcasts poses a significant challenge due to the need for a massive database of faces, high processing speed, and precision, especially when dealing with similar celebrity faces.

Innovation Solution

The disclosed architecture enables real-time entity recognition, including face recognition, by extracting live data, filtering named entities, and using a trained model to recognize faces in video frames, even with modest resources, and supplements the video with relevant content from various sources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a massive database of faces is created to improve recognition precision, then measurement precision is improved, but device complexity and memory resources increase significantly

Engineering Contradiction:
Improveface recognition precisionVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the face recognition system into multiple components: face detection module, feature extraction module, and recognition module. The database is segmented into structured fields (identity, biographical information, contact information) rather than storing complete images. This segmentation reduces memory requirements while maintaining recognition precision by processing only essential features.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts only the essential facial features and identity information needed for recognition, rather than storing and processing complete face images. The system extracts feature vectors and key identifiers from faces, removing redundant visual data while preserving recognition capability. This extraction approach reduces database size and processing complexity.

Inventive Principle:
Principle #2Taking out (Extraction)

2Productivity

If real-time processing is implemented to improve speed, then productivity is improved, but measurement precision deteriorates due to processing limitations

Engineering Contradiction:
Improveprocessing speedVSAvoidface recognition precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent performs preliminary face detection and feature extraction before the actual recognition process. Detection frames are pre-processed to extract facial features and create feature vectors in advance. This preliminary action prepares the data for rapid comparison during real-time recognition, maintaining both speed and precision by reducing the computational burden during the critical recognition moment.

Inventive Principle:
Principle #10Preliminary action

3Device complexity

If a smaller subset of faces is processed to reduce memory requirements, then device complexity is reduced, but measurement precision worsens due to limited database coverage

Engineering Contradiction:
Improvememory resourcesVSAvoidface recognition precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by storing complete information for a smaller subset of recognized faces rather than attempting to store compressed data for a larger population. Each recognized face in the database maintains high-quality, complete attribute records. This approach achieves acceptable precision for targeted individuals while keeping memory requirements manageable by focusing resources on a curated subset.

Inventive Principle:
Principle #3Local quality

4Measurement precision

If similar celebrity faces are distinguished to improve measurement precision, then measurement precision is improved, but productivity deteriorates due to increased processing requirements

Engineering Contradiction:
Improveface distinction precisionVSAvoidprocessing speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent changes the parameter representation from complete facial images to condensed feature vectors with key distinguishing attributes. By transforming the data representation into a more efficient parameter space, the system can distinguish between similar faces using reduced-dimensional feature comparisons rather than full-image processing. This parameter transformation maintains discrimination precision while dramatically reducing processing requirements for real-time operation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3311332B1Automatic recognition of entities in media-captured events
Publication Date: 2025.01.29 MICROSOFT TECHNOLOGY LICENSING LLC
  • EP3311332B1 patent drawingFigure 1
  • EP3311332B1 patent drawingFigure 2
  • EP3311332B1 patent drawingFigure 3

AI summary

Architecture that enables the identification of entities such as people and content in live broadcasts (e.g., streaming content (e.g., video) of live events) and non-live presentations (e.g., movies), in realtime, using recognition processes. This can be accomplished by extracting live data related to a live event. With respect to people entities, filtering can be performed to identify the named (people) entities from the extracted live data, and trending topics discovered as relate to the named entities, as associated with the live event. Multiple images of the named entities that capture the named entities under different conditions are captured for the named entities. The images are then processed to extract and learn facial features (train one or more models), and facial recognition is then performed on faces in the video using the trained model(s).