Biometric Object Detection Using Segmented Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition technologies, such as the YOLO technology, are not efficiently adapted for identifying specific biometric objects like fingertips, requiring longer processing times and significant computer resources, making them unsuitable for current mobile devices and sensitive applications like user identification for bank accounts.
Innovation Solution
A method implemented on a mobile device that uses a neural network to process images of biometric objects, including a liveliness-detection operation to distinguish between real and spoof objects, and a comparison operation to determine the authenticity of the object, thereby reducing the required computer resources and enabling real-time identification on mobile devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If YOLO technology is used for identifying biometric objects like fingertips, then real-time detection capability is improved, but detection accuracy for specific biometric objects deteriorates
Solution Approach 1:
The patent segments the identification task into two distinct stages: a first neural network performs initial object detection and classification, while a second specialized neural network focuses specifically on biometric characteristic identification. This segmentation allows each network to be optimized for its specific function, maintaining real-time performance while improving accuracy for biometric objects like fingertips.
Solution Approach 2:
The patent applies local quality by using a second neural network specifically tailored for biometric object identification after the initial detection. This specialized network processes only the regions of interest identified by the first network, applying enhanced processing quality specifically where needed to improve detection accuracy for biometric characteristics without compromising overall real-time performance.
2Measurement precision
If complex neural networks are used for object identification, then identification accuracy is improved, but required computer resources worsen
Solution Approach 1:
The patent divides the identification system into two neural networks with distinct functions. The first network handles general object detection and classification, while the second network specialized in biometric characteristics. This segmentation reduces the computational burden on any single network, allowing each to be smaller and more efficient while maintaining overall high accuracy.
Solution Approach 2:
The patent applies partial action by having the second neural network process only the specific biometric regions of interest identified by the first network, rather than processing the entire image. This selective processing reduces the computational resources required while maintaining high identification accuracy for biometric characteristics.
3Adaptability or versatility
If existing image recognition technologies are used for user identification, then general object recognition capability is improved, but suitability for sensitive applications like bank accounts deteriorates
Solution Approach 1:
The patent segments the recognition process into two stages with distinct purposes: the first network handles general object recognition for versatility, while the second network provides specialized biometric verification for security. This segmentation enables the system to maintain both general adaptability and high reliability for sensitive applications like bank account authentication.
Solution Approach 2:
The patent introduces an intermediary mechanism where the first neural network's output serves as input to the second neural network. This intermediary processing stage ensures that general object recognition is followed by specialized biometric verification, adding a layer of security that makes the system suitable for sensitive applications while preserving general recognition capabilities.
Data Source
AI summary
A method includes obtaining, by a processing device from an optical sensor of a mobile device, an image of an object. The method further includes processing, by the processing device, the image to identify the object in the image. Processing the image includes calculating at least one of a distance map or a reflection pattern. Processing the image further includes comparing the calculated distance map or the calculated reflection pattern with a known distance map or a known reflection pattern to determine whether the image contains a spoof or a real object. The method further includes obtaining, by the processing device from the image, data including a biometric characteristic of the user. The method further includes sending, by the processing device to a third party computing device, data including at least the biometric characteristic.