Facial Recognition via High Probability Group Forecasting

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Facial recognition in edge and near-edge computing environments faces challenges due to limited computational power, connectivity latency, and low recognition rates for unseen angles, particularly in real-time applications.

Innovation Solution

The implementation of a High Probability Group (HPG) approach that predicts and prioritizes faces likely to appear in a video stream based on historical data, using a forecasting procedure to augment a subset database with relevant faces, allowing for efficient face recognition and reduced latency by operating on a smaller dataset.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a large database of faces is employed in facial recognition processes, then recognition accuracy is improved, but processing time increases significantly

Engineering Contradiction:

Improverecognition accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by predicting which faces are likely to appear in upcoming video frames based on historical appearance patterns. Faces are pre-sorted and prioritized into a reduced search database before actual recognition occurs, so that when faces appear, they can be quickly identified without searching the entire database. This predictive pre-sorting resolves the contradiction by preparing the search space in advance based on temporal patterns.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The full face database is segmented into multiple subsets based on predicted appearance probability and temporal patterns. Instead of searching the entire database, the system divides faces into high-probability groups that are more likely to appear in upcoming frames. This segmentation reduces the effective search space while maintaining accuracy for expected appearances, resolving the time-accuracy tradeoff.

Inventive Principle:

Principle #1Segmentation

2Speed

If facial recognition is performed in edge computing environments, then real-time processing capability is improved, but computational power constraints worsen performance

Engineering Contradiction:

Improvereal-time processing capabilityVSAvoidcomputational power

Core Design Contradiction:

SpeedVSPower

Solution Approach 1:

The system applies partial action by using a reduced subset of the face database for actual recognition operations. Instead of processing all faces in the database, the system selectively processes only the predicted high-probability faces that are likely to appear. This partial processing approach enables real-time performance on edge devices with limited computational power while maintaining acceptable accuracy for expected appearances.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The system performs preliminary prediction of face appearances and pre-organizes the search database before actual recognition. This pre-computation of appearance probabilities and database sorting reduces the computational burden during real-time processing, enabling edge devices to achieve real-time performance without requiring full database search capability.

Inventive Principle:

Principle #10Preliminary action

3Power

If cloud connectivity is used for facial recognition, then computational resources are improved, but connectivity latency increases processing time

Engineering Contradiction:

Improvecomputational resourcesVSAvoidconnectivity latency

Core Design Contradiction:

PowerVSLoss of time

Solution Approach 1:

The system performs all database sorting, face prediction, and search optimization locally on edge devices before needing cloud connectivity. By pre-computing appearance probabilities and organizing the reduced search database locally, the system minimizes or eliminates the need for continuous cloud connectivity during real-time processing, resolving the latency issue while still benefiting from cloud resources for model training and updates.

Inventive Principle:

Principle #10Preliminary action

4Adaptability or versatility

If faces are captured at unseen angles, then environmental adaptability is improved, but recognition rate decreases

Engineering Contradiction:

Improveenvironmental adaptabilityVSAvoidrecognition rate

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The system dynamically adapts to different capture angles by continuously learning and updating appearance patterns from multi-angle video data. The predictive model adjusts its face appearance expectations based on the actual angles captured in the environment, allowing the system to maintain high recognition rates across varying perspectives. This dynamic adaptation resolves the contradiction by making the recognition system flexible to environmental variations.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS11954941B2Facial recognition IOT application for multi-stream video using forecasting tools

Publication Date: 2024.04.09 EMC IP HLDG CO LLC

AI summary

One example method includes accessing an appearance history of a person, the appearance history including information concerning an appearance of a person at a particular time, generating, based on the appearance history, a forecast that comprises a probability that the person will appear again at some future point in time, determining that the probability meets or exceeds a threshold, and updating a high probability group database to include a facial image of a face of the person.