Edge Facial Recognition Bandwidth Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facial recognition systems in cameras face inefficiencies due to limited uplink bandwidth in wireless networks, leading to delayed image transmission and excessive battery consumption when high-resolution images are sent to backend systems for analysis, while downlink bandwidth often remains underutilized.
Innovation Solution
The method involves capturing images at a high resolution on edge devices and converting them to a lower resolution based on available uplink and downlink bandwidth, allowing for efficient transmission of reduced data sets, which are then used for local facial recognition, reducing processing power and storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution images are transmitted over uplink bandwidth, then facial recognition accuracy is improved, but bandwidth consumption increases and transmission time increases
Solution Approach 1:
The patent segments the facial recognition process into two stages: first transmitting a low-resolution image over uplink for initial candidate identification, then transmitting only the candidate facial templates over downlink for final verification. This segmentation allows high-resolution data to be processed in stages, reducing uplink bandwidth consumption while maintaining recognition accuracy.
Solution Approach 2:
The patent exploits the asymmetric dimension of wireless bandwidth by utilizing both uplink and downlink channels with different capacity characteristics. Low-resolution images are sent over the constrained uplink, while candidate templates are received over the higher-capacity downlink, effectively using another dimension (downlink bandwidth) to compensate for uplink limitations.
2Measurement precision
If high-resolution images are transmitted over uplink bandwidth, then facial recognition accuracy is improved, but transmission time increases
Solution Approach 1:
The patent extracts only the essential information needed for initial matching by converting high-resolution images to low-resolution versions for uplink transmission. This extraction reduces the data volume significantly, enabling faster transmission while still capturing sufficient facial features for candidate identification, thereby reducing transmission time without completely sacrificing accuracy.
Solution Approach 2:
The system performs preliminary facial recognition using low-resolution images transmitted quickly over uplink to identify candidate matches. This preliminary action filters down the candidate set before receiving full-resolution templates over downlink, reducing the overall processing time by quickly eliminating non-matching candidates without waiting for large high-resolution image transfers.
3Measurement precision
If high-resolution images are processed locally, then facial recognition accuracy is improved, but processing power and storage needs increase
Solution Approach 1:
The patent introduces a backend server as an intermediary that stores the comprehensive facial template database. Instead of requiring the edge device to store and process all possible facial templates locally, the edge device only needs to store and process the much smaller set of candidate templates received from the backend, significantly reducing local processing power and storage requirements while maintaining recognition accuracy.
4Productivity
If downlink bandwidth is underutilized, then network resource efficiency decreases, but uplink bandwidth remains limited
Solution Approach 1:
The patent inverts the traditional approach by sending low-resolution images uplink and receiving candidate templates downlink, rather than sending all high-resolution images uplink. This inversion leverages the typically higher capacity of downlink bandwidth to transfer the larger candidate template data, thereby utilizing downlink resources more efficiently while working within uplink bandwidth constraints.
Data Source
AI summary
Techniques for leveraging downlink bandwidth when uplink bandwidth is limited are provided. An image is captured at an edged device, the image including at least one face of a person, the image captured at a first resolution. The image is stored at the first resolution in the edge device. The image is converted to a second resolution, the second resolution being lower than the first resolution. The converted image is sent to a backend facial recognition system. A set of candidate facial recognition matches is received. Facial recognition is performed at the edge device based on the stored image captured at the first resolution and the set of candidate facial recognition matches.