Live Broadcast User Identification via Two-Stage Object Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in efficiently identifying individuals in live event broadcasts and generating notifications to connected devices, particularly due to the high computational resources required for facial recognition in large crowds.

Innovation Solution

A system that generates a link between a user and an event, allowing for efficient identification of users in event captures through facial recognition or biometric means, and subsequently transmitting notifications to associated devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If facial recognition is performed on all frames of a live broadcast capture, then user identification accuracy is improved, but computational resource consumption increases significantly

Engineering Contradiction:

Improveuser identification accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system segments the large-scale facial recognition problem into two stages: first, identify candidate regions using object detection on selected frames; second, perform facial recognition only on these candidates. This segmentation reduces the computational burden while maintaining identification accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary object detection to identify candidate regions before conducting facial recognition. This preliminary action filters out non-relevant areas, so that subsequent facial recognition is applied only to promising candidates, reducing overall computational resources needed.

Inventive Principle:

Principle #10Preliminary action

2Reliability

If facial recognition is performed on every frame of the capture, then user identification reliability is improved, but processing time increases

Engineering Contradiction:

Improveuser identification reliabilityVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs facial recognition periodically on selected frames rather than continuously on every frame. By strategically choosing which frames to analyze based on object detection results, the system maintains identification reliability while reducing processing time.

Inventive Principle:

Principle #19Periodic action

Solution Approach 2:

Object detection is performed as a preliminary action to identify frames containing potential targets before conducting facial recognition. This preliminary filtering ensures that facial recognition is applied only when necessary, reducing processing time while maintaining reliability.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If the system analyzes all captured material from multiple image capture devices, then user identification completeness is improved, but device complexity increases

Engineering Contradiction:

Improveuser identification completenessVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system segments the analysis task by device and frame, processing each capture device's output independently through the object detection and facial recognition pipeline. This modular segmentation manages system complexity while ensuring comprehensive user identification across all devices.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Object detection serves as an intermediary step between raw capture data from multiple devices and the final facial recognition process. This intermediary filters and prioritizes data, making the overall system more manageable while maintaining completeness of user identification.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Measurement precision

If the system processes high-resolution capture data from live broadcasts, then user identification precision is improved, but computational resource consumption increases

Engineering Contradiction:

Improveuser identification precisionVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The system performs preliminary object detection on high-resolution capture data to identify candidate regions before applying computationally intensive facial recognition algorithms. This preliminary action reduces the amount of high-resolution data that requires full processing, lowering computational resource consumption while maintaining identification precision.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250071370A1Systems and methods for generating a notification in response to a live broadcast

Publication Date: 2025.02.27 ADEIA GUIDES INC

AI summary

Systems and methods are provided for generating a notification in response to identifying a user in a capture of an event. A link between a user and an event is generated and a capture of the event is received at a first computing device. The user in the capture of the event is identified at the computing device and based on the link. A notification is generated based on identifying the user. The notification is transmitted to a second computing device.