Skin Hue Correction for 1-Second Liveness Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric authentication systems face challenges in accurately distinguishing between living and non-living bodies within a short time frame due to the mixing of pulsation and noise frequency components, leading to prolonged determination times and user inconvenience.
Innovation Solution
A video processing apparatus that corrects skin color hues in video frames using a correction coefficient and determines living bodies based on average hue values, employing a discriminator to analyze amplitude ratios of corrected hues for accurate classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If FFT is used to acquire frequency spectrum from time-series color data, then frequency spectrum can be obtained, but determination time becomes excessively long requiring 3-5 seconds of data
Solution Approach 1:
The patent extracts only the necessary frequency information (pulsation frequency component) from the color signal using band-pass filtering, rather than performing a complete FFT analysis of the entire frequency spectrum. This extraction approach obtains the critical authentication information without the computational overhead of full spectrum analysis, enabling rapid determination while maintaining accuracy.
Solution Approach 2:
The patent applies band-pass filtering as a preliminary processing step to isolate the pulsation frequency band before further analysis. This preliminary action pre-separates the relevant frequency components, eliminating the need for lengthy FFT processing and enabling quick authentication decisions within 1 second or less.
2Measurement precision
If time-series data of 3-5 seconds is used to distinguish pulsation frequency from noise, then frequency components can be distinguished, but user convenience deteriorates due to prolonged waiting time
Solution Approach 1:
The patent extracts and isolates only the pulsation frequency band using band-pass filtering, removing the need to analyze the entire frequency spectrum. This extraction enables accurate pulsation detection from short 1-second videos by focusing computational resources on the relevant frequency range, thereby improving user convenience while maintaining measurement precision.
Solution Approach 2:
Instead of performing complete spectrum analysis, the patent applies partial action by using band-pass filtering to target only the specific pulsation frequency range. This selective approach achieves sufficient authentication accuracy with significantly reduced processing time, allowing authentication within 1 second and greatly improving user convenience.
3Productivity
If short video of about one second is used for authentication, then user wait time is reduced, but frequency components of pulsation and noise become mixed making accurate determination difficult
Solution Approach 1:
The patent introduces band-pass filtering as an intermediary processing step that separates pulsation frequency components from noise. This intermediary filter acts as a mediator between the short video input and the authentication decision, enabling accurate spoofing determination from 1-second videos by selectively passing only the relevant pulsation frequencies while blocking noise.
Solution Approach 2:
The patent changes the processing parameter from complete FFT spectrum analysis to band-pass filtered frequency extraction. This parameter change transforms the approach from analyzing all frequency components to selectively extracting only the pulsation band, enabling accurate authentication from short 1-second videos and significantly improving productivity.
Data Source
AI summary
The present invention provides a technique for determining whether an object is a living body with a high degree of accuracy even for a relatively short video of about one second. To achieve this, a video processing apparatus configured to determine whether an object is a living body from video data of the object constituted by a plurality of frames, comprises a correction unit configured to correct each of the plurality of frames constituting the video data based on a correction coefficient for correcting to colors of a plurality of target hues different from each other; and a determination unit configured to determine whether the object is a living body based on an average hue value of a skin area of the object in each of the plurality of frames being corrected.