Multi-Spectral Liveness Detection Using Neural Network Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition systems for authentication struggle to differentiate between a live human face and a spoofed face, particularly with two-dimensional representations, leading to delays and inaccuracies in liveness detection.
Innovation Solution
A system that captures two-dimensional image data in multiple spectral bands, creates a combined image representation, and uses a neural network to classify the data as depicting a live human person or a spoof, enabling real-time liveness detection by analyzing for implied three-dimensional features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple frames of video stream or static images captured at multiple frame times are used for liveness detection, then liveness detection accuracy is improved, but processing time increases and real-time detection is compromised
Solution Approach 1:
The patent transitions from temporal analysis (multiple frames over time) to spectral analysis (multiple wavelengths simultaneously). By capturing images at different wavelengths in a single frame and analyzing spectral reflectance characteristics, the system achieves accurate liveness detection without the time delays associated with collecting multiple temporal frames.
Solution Approach 2:
The system changes the parameter of image capture from temporal sequencing to spectral variation. Instead of capturing multiple frames at different times, the patent captures a single frame with multi-spectral information, analyzing wavelength-dependent reflectance properties to detect liveness, thereby reducing processing time while maintaining accuracy.
2Measurement precision
If three-dimensional cameras are used to create point cloud maps for liveness detection, then liveness detection accuracy is improved, but device complexity and processing requirements increase
Solution Approach 1:
The patent uses a two-dimensional image sensor to capture spectral information that effectively represents three-dimensional surface properties through wavelength-dependent reflectance analysis. This approach creates a spectral fingerprint copy of the surface characteristics without requiring actual three-dimensional camera hardware, thereby reducing device complexity while maintaining detection accuracy.
Solution Approach 2:
The system replaces the mechanical complexity of three-dimensional camera hardware and point cloud generation with a simpler spectral analysis approach. By using a standard 2D sensor with multi-spectral filtering and analyzing reflectance patterns across wavelengths, the patent achieves equivalent liveness detection without the complex optical systems and extensive post-processing required by 3D cameras.
3Productivity
If spectral reflectance analysis at multiple wavelengths is performed on a single image, then real-time liveness detection is achieved, but measurement complexity increases
Solution Approach 1:
The patent segments the spectral analysis into distinct wavelength bands that can be processed independently. By dividing the multi-spectral data into separate wavelength ranges and analyzing reflectance characteristics in each segment, the system simplifies the overall measurement complexity while maintaining the ability to detect liveness in real-time from a single captured image.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The system provides improved classification performance and real-time liveness detection, effectively preventing unauthorized access using two-dimensional spoofs, with enhanced accuracy and reduced processing time compared to existing methods.
Implementation Method 1
capturing, by the image capture device, respective two-dimensional image data representing two images of the human person in two spectral bands
Data Source
AI summary
Systems for detecting liveness in image data may perform a process including receiving, from one or more image capture devices, two-dimensional image data representing two images captured simultaneously and depicting the same human person, including data representing light captured in two spectral bands, such as visible light and infrared light. The received image data may be converted to grayscale or downsampled prior to further processing. The process may include creating a two-dimensional combined image representation of the received image data and analyzing the combined image representation to detect any implied three-dimensional features using a neural network or machine learning. The process may include classifying the received image data as likely to depict a live person or a two-dimensional spoof of a live person, dependent on the analysis, and outputting a classification result indicating a likelihood that the received image data depicts a live person.