Biometrics in Artificial Intelligence Applications

liu, tempo Date: 2021-09-06 10:37:55
Views:83 Reply:0

Bone Voice Recognition

Yes, it’s bone voice recognition, not voice recognition. Bone voiceprint is the use of headphones to collect the echo of each person’s unique voice reflected in the transmission between their unique bones, relying on the two major technologies of bone conduction forensics and voice print forensics to process the voiceprint transmitted to the cell phone, so that bone voiceprint cannot be cracked by recording the voice, so the security performance is higher.


This is a relatively remote term, and there are already commercial cases at the moment. For example, the FlyPods Pro headset equipped with the Honor phone is equipped with bone acoustic pattern recognition technology, which can realize the functions of unlocking the phone screen and calling up the payment QR code page. Users need to wear FlyPods Pro left and right headphones, and in a quiet environment according to the guidelines to record bone voice pattern.


Specifically, when the user wears the headphones, one side of the device will play a specific band of audio that is imperceptible to the human ear for a very short period of time and is harmless to humans. The other side of the headset will temporarily act as a sound wave receiving device that will recognize, identify and record the sound through the skull.


As each person’s skull structure is unique, so the reflected sound is also unique. The system will then analyze the voice pattern, detect and match it with the database to achieve the purpose of securely unlocking mobile devices. This technology is currently only used by Huawei, so there is no targeted cracking method.


Cochlear recognition

Back in 2016, the Japanese company NEC introduced a new way of unlocking – cochlear implant verification unlock.


NEC created a miniature in-ear earbud that can measure the way the cochlea vibrates after making a sound for authentication. Each person has a unique construction pattern for the shape of the ear, and NEC says this identification method is 99% accurate and takes only 1 second to complete.


Artificial Intelligence Applications


NEC’s experimental results show that both the eardrum and the inner ear canal respond to the transmission of sound signals as samples for individual differences verification, and that the signal eigenvalues reduce the complexity of identification calculations while increasing the speed and accuracy.


NEC intends to focus on fraud prevention and protection in the areas of personal and property security, or the maintenance and management of security infrastructures such as healthcare and security. At present, there is no news about the cracking of cochlear recognition as it is not found to be commercially available on a large scale.


Signature Recognition

Signature recognition has been around since ancient times and is still used today, for example, when signing contracts and agreements. Each person has his or her own unique writing style, so signature recognition can be used as a type of biometric identification and is a behavioral measurement. Signature recognition has a digital process: the measurement of the image itself as well as the whole action of signing, the different speed, sequence and pressure in each character and between characters.


Signature recognition is divided into offline signature recognition and online signature recognition.


Online recognition is to collect the signature sample of the writer through the handwriting board, and some systems can also collect data such as pressure and the angle of holding the pen, and the current recognition rate is quite high. However, since human writing action is not fixed, it may not be possible to distinguish real signature and forged signature from glyphs alone. Therefore, offline signatures are more insecure.


In ancient times, there were simulated glyphs to forge certificates and even imitation of holy decrees. Today, signature recognition has become an auxiliary authentication, not unique, so even if the signature recognition is cracked, it often does not make much sense.


Endnote: Brainstorm series

In addition, some readers mentioned memory recognition.


Isn’t memory recognition simply a matter of remembering passwords or secrets? But the reader dispersed more scenarios, such as dynamic questions based on your experience at any time, and so on. After all, people think that memory can’t be cracked. Can memory be cracked? At the moment, it seems that indeed it cannot. Because the formation of human self-awareness, the synthesis of the main self-perception, self-memory and self-thought, of which memory occupies a considerable weight.


If memory can be deciphered, then it should not be difficult to copy it. Once memories can be copied, it is possible to create an infinite number of “I’s”, which seems to easily cause confusion in the ethical and philosophical system, and to develop in a completely unpredictable direction.


At this point, the brain may not be quite enough. So I leave it to you, the reader, to develop your thinking later on. Even if memory can become a biometric, then the consequences will be unimaginable. Because, memory is easy to forget.


In fact, combing some found that there is really no one biometric, is infallible security. Although security is not absolute, but we only need relative security. After all, as we ordinary people, the black and gray industry also do not need to invest a lot of money and energy for you, just to crack your phone. The future development trend of security verification should be the biometric technology of multiple checks, especially the interactive, AI intelligent logic judgment biometric.

Leave a comment

You must Register or Login to post a comment.