Deception Detection
Through a multimodal video stream processing system, combined with deep learning technology, real-time analysis of facial biometrics, the accuracy and reliability problems of long-distance contactless fraud detection and remote physiological monitoring in the existing technology are solved, and high-precision detection and monitoring effects are achieved.
Patent Information
- Application Number
- JP2024551908
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-02-28
- Filing Date
- 2023-02-28
- Publication Date
- 2025-03-21
AI Technical Summary
The prior art is difficult to achieve high-precision long-distance contactless fraud detection and remote physiological monitoring, especially under multimodal conditions, and is greatly affected by environmental factors and subject motion.
Through a multimodal video stream processing system, combining visible light, near-infrared, far-infrared and thermal video streams, facial biometrics, such as pulse rate, pupil diameter, facial temperature, speech and micro-expression, it uses deep learning technologies such as 3D convolutional neural network (3DCNN) for signal processing and feature fusion to achieve fraud detection and physiological monitoring.
It realizes high-precision long-distance non-contact fraud detection and remote physiological monitoring without restricting subject movement and posture, improving the accuracy and reliability of the detection.