Synthesized Video Detection via Blood Flow Signal Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting synthesized videos, such as DeepFakes, are inefficient, inaccurate, and require frequent updates due to their reliance on deep learning models that analyze video features like resolution and frame rate, making them computationally intensive and slow.
Innovation Solution
The system employs remote video photoplethysmography to capture light re-emitted from the skin, processing blood flow signals using machine learning models trained on hemoglobin concentration changes and blood flow data to identify abnormalities indicative of synthesized videos, which are less computationally resource-intensive and faster.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deep learning models are used to analyze video features for detecting synthesized videos, then detection capability is provided, but computational intensity increases and processing speed decreases
Solution Approach 1:
The patent extracts and analyzes only the blood flow signal component from the video, rather than processing the entire video frame by frame. By isolating the physiological signal related to hemoglobin concentration changes, the system reduces the computational burden while maintaining detection capability for synthesized videos.
Solution Approach 2:
The patent replaces complex deep learning visual analysis with a physiological signal-based detection method. Instead of using computationally intensive neural networks to analyze video pixels, the system uses machine learning models trained on blood flow signal patterns, which are less computationally demanding and faster to process.
2Measurement precision
If deep learning models analyze video features for synthesized video detection, then detection accuracy is achieved, but computational resource consumption increases
Solution Approach 1:
The system extracts only the relevant blood flow signal information from the video data, discarding unnecessary visual information. This selective extraction reduces the amount of data that needs to be processed computationally while maintaining sufficient accuracy for detection purposes.
Solution Approach 2:
The patent transforms the detection approach from analyzing visual parameters (resolution, frame rate) to analyzing physiological parameters (blood flow signals, hemoglobin concentration changes). This parameter transformation enables more efficient computational processing while maintaining detection accuracy.
3Reliability
If existing detection methods are updated frequently to maintain accuracy, then detection reliability is maintained, but system complexity and maintenance burden increase
Solution Approach 1:
The patent employs a detection approach that uses readily available physiological signal data and established machine learning models, rather than relying on complex, constantly evolving deep learning systems. This reduces the need for frequent model updates and system retraining, lowering maintenance complexity while maintaining reliability.
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
This approach allows for rapid and accurate identification of synthesized videos by analyzing blood flow patterns, reducing computational load and improving detection speed compared to existing methods.
Implementation Method 1
The system employs remote video photoplethysmography to capture light re-emitted from the skin
Data Source
AI summary
A system and method for detection of synthesized videos of humans. The method including: determining blood flow signals using a first machine learning model trained with a hemoglobin concentration (HC) changes training set, the first machine learning model taking as input bit values from a set of bitplanes in a captured image sequence, the HC changes training set including bit values from each bitplane of images captured from a set of subjects for which HC changes are known; determining whether blood flow patterns from the video are indicative of a synthesized video using a second machine learning model, the second machine learning model taking as input the blood flow signals, the second machine learning model trained using a blood flow training set including blood flow data signals from at least one of a plurality of videos of other human subjects for which it is known whether each video is synthesized.