| Abstract |
Fingerprint scanning is a biometric technology widely used in various fields, such as mobile payment and intelligent security, due to its high accuracy and reliability. Compared with capacitive and optical fingerprint scanning, ultrasonic fingerprint scanning can fix the authentication issue when fingertips are wet, oily or dusty. However, achieving fast and accurate ultrasonic fingerprint scanning and authentication in a large physical domain is challenging. A novel technique, combining deep learning and tomography, is suggested by this study to tackle these matters. In this study, fingerprint reconstruction at any position is achieved in a large physical domain by monitoring the wave field changes of a plate-like structure using ultrasonic guided waves. Accurately characterization and authentication of the fingerprint can be achieved through the utilization of fast inversion tomography (FIT) and mask region convolutional neural network (Mask R-CNN). A sequence of data and image processing shows that FIT can effectively characterize fingerprint features in the array, and the post-processing method based on the Mask R-CNN model can achieve fast and accurate ultrasonic fingerprint matching in a large physical domain. This further enhances the comfort and security of devices that use fingerprint identification. |
| Authors |
Shuainan Chen , Chengwei Zhao  , Jian Li , Min Lin  , Zebing Zeng , Liu Yang
|
| Journal Info |
Not listed | 2023 IEEE International Ultrasonics Symposium (IUS)
|
| Publication Date |
9/3/2023 |
| ISSN |
Not listed |
Type |
article |
| Open Access |
closed
|
| DOI |
https://doi.org/10.1109/ius51837.2023.10306484 |
| Keywords |
Fingerprint (Score: 0.572829) , Multimodal Biometrics (Score: 0.532157) , Authentication (Score: 0.513114) , Fingermark Residue (Score: 0.511557) , Face Spoof Detection (Score: 0.506875)
|
