Enhanced dorsal hand vein recognition using convolutional neural network with long-short term memory model
Abstract
Recently, biometrics technology has become increasingly vital for personal authentication. This study sought to develop a robust identification system based on hand-vein images using a CNN Model augmented with LSTM. To enhance recognition accuracy, the study proposed implementing various preprocessing filters. The involved dataset included 500 hand-vein images from 50 patients, and the training was implemented by data augmentation. The model achieved a 70% classification accuracy without image processing. Additionally, the mean filter was applied to eliminate noise and achieved an impressive accuracy of 99% in both training conditions.
Fields of Study & Subjects
Biometric Identification and Security
Handwritten Text Recognition Techniques
Hand Gesture Recognition Systems
Publication details
- Journal / source
- AIP conference proceedings
- Volume
- 3406
- Pages
- 020060–020060
- Publication date
- 1 January 2026
- Language
- EN