Deep Neural Network Based on A Stacked Autoencoder For 3D-Finger Knuckle Recognition

The recognition of 3D_finger knuckle is unique to the encouraging biometric techniques, which has usual great interest recently due to its accuracy in recognizing individuals Though the literature observed several techniques and progresses to deal with the problem of identifying persons through the...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2023 International Conference on Information Technology, Applied Mathematics and Statistics (ICITAMS) s. 140 - 144
Hlavní autoři: Al-Janabi, Dua'a Hamed, Al-Juboori, Ali Mohsin
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 20.03.2023
Témata:
3D_finger knuckle
autoencoder
biometric
Biometrics (access control)
deep learning
Fingers
Image recognition
Softmax classification
Solid modeling
stack sparse autoencoders(SAEs)
Text recognition
Three-dimensional displays
Training
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:The recognition of 3D_finger knuckle is unique to the encouraging biometric techniques, which has usual great interest recently due to its accuracy in recognizing individuals Though the literature observed several techniques and progresses to deal with the problem of identifying persons through the finger knuckle, the tools are still in its beginning. This paper offers an effective deep learning technique, stacked sparse autoencoders (SSAEs), to identify the individual through a 3d finger knuckle. The suggested method can extract features from finger knuckles and remove the exhausting use of features. Finally, a softmax classification model was introduced into stacked SAEs to classify a set of features. The suggested model was developed using three dimensions (3D) Finger Knuckle Images datasets from The Hong Kong Polytechnic University (PolyU);. The SSAE learns high-level features from just pixel intensities alone to identify distinguishing features of 3d finger knuckle images. Stacked autoencoders another better way for the feature extraction SSAE was shown to have an improved accuracy of 99%.
DOI:10.1109/ICITAMS57610.2023.10525418