Researchers in India have created a unique database of 8,000 hand vein and palmprint images using newly created hardware that captures four biometrics from each hand scanned.
According to a paper from the PES College of Engineering, from each hand, the multimodal device can record the palmprint, dorsal vein, wrist vein and palm vein. Contact is required.
The hardware used near infra-red light-emitting diodes to image to illuminate the veins and a single-camera CMOS sensor to record them. Palmprints were lit using multiple white LEDs.
The researchers scanned from 308 people to get 8,336 hand vein and palm print images from two hands each. Almost half of the participants were 20 to 30 years old. About a third were 15 to 20 years old. The rest were 30 to 80 years old. The creation of new biometric datasets has been a regular source of ethical and privacy concerns.
Classification accuracy with the multi-modal approach was 99.8 percent in k nearest neighbor, 99 percent in random forest, 99.8 percent in support vector machine, 99.8 percent in neural network, 97.7 percent in gradient boosting and 99.8 in percent logistic regression.
Multi-modal biometric systems to date have generally referred to systems that combine hand- or fingerprints with facial or iris recognition.
Fujitsu Laboratories in January 2021 showed off a multi-modal, contactless ID authentication retail gait. The system performed palm recognition and facial verification with or without a mask.
A month later, the U.S. Navy discussed a system it was using that collects face, fingerprint and iris biometric identifiers.
Researchers in India have created a unique database of 8,000 hand vein and palmprint images using newly created hardware that captures four biometrics from each hand scanned.
According to a paper from the PES College of Engineering, from each hand, the multimodal device can record the palmprint, dorsal vein, wrist vein and palm vein. Contact is required.
The hardware used near infra-red light-emitting diodes to image to illuminate the veins and a single-camera CMOS sensor to record them. Palmprints were lit using multiple white LEDs.
The researchers scanned from 308 people to get 8,336 hand vein and palm print images from two hands each. Almost half of the participants were 20 to 30 years old. About a third were 15 to 20 years old. The rest were 30 to 80 years old. The creation of new biometric datasets has been a regular source of ethical and privacy concerns.
Classification accuracy with the multi-modal approach was 99.8 percent in k nearest neighbor, 99 percent in random forest, 99.8 percent in support vector machine, 99.8 percent in neural network, 97.7 percent in gradient boosting and 99.8 in percent logistic regression.
Multi-modal biometric systems to date have generally referred to systems that combine hand- or fingerprints with facial or iris recognition.
Fujitsu Laboratories in January 2021 showed off a multi-modal, contactless ID authentication retail gait. The system performed palm recognition and facial verification with or without a mask.
A month later, the U.S. Navy discussed a system it was using that collects face, fingerprint and iris biometric identifiers.