Method and apparatus for key generation based on face recognition using CNN and RNN
A face recognition based key generation apparatus controls a key generation model that is formed of a CNN and an RNN to be learned to generate a desired key having a consistent value by using sample facial images of a key owner and a PIN of the key owner as inputs, and the key generation model receives a facial image of the key owner and the PIN of the key owner, as inputs at a desired key generation time, and generates a key.
1. A face recognition based key generation method comprising:
learning a key generation model that is formed of a convolutional neural network (CNN) and a recurrent neural network (RNN) for outputting a consistent key value by using sample facial images of a key owner and a personal identification number (PIN) of the key owner as inputs; and
generating a key by receiving a facial image of the key owner and the PIN of the key owner at a desired key generation time, by the key generation model,
wherein the key is for performing a digital signature, and
wherein the learning the key generation model comprises:
outputting a feature value of the respective sample facial images, which are provided as inputs, from the CNN, and
performing iterative learning to output the consistent key using the feature value of each sample facial image and the PIN as inputs, in the RNN.
2. The face recognition based key generation method of claim 1 , wherein the CNN is learned to generate a feature value from each sample facial image.
3. The face recognition based key generation method of claim 1 , further comprising, when an error correcting code is included in the key output from the key generation model, correcting the output key value and eliminating the error correcting code.
4. A face recognition based key generation apparatus comprising:
a key generation model formed of a convolutional neural network (CNN) and a recurrent neural network (RNN), the key generation model receiving a facial image of a key owner and a personal identification number (PIN) of the key owner, as inputs at a desired key generation time, and generating a key; and
a processor that controls the key generation model to be learned to generate a desired key having a consistent value by using sample facial images of a key owner and a PIN of the key owner as inputs,
wherein the key is for performing a digital signature, and
wherein the learning the key generation model comprises:
outputting a feature value of the respective sample facial images, which are provided as inputs, from the CNN, and
performing iterative learning to output the consistent key using the feature value of each sample facial image and the PIN as inputs, in the RNN.
5. The face recognition based key generation apparatus of claim 4 , further comprising a function block that corrects an error in a key value output from the key generation model and eliminates an error correcting code when the error correcting code is included in the key value.
6. The face recognition based key generation apparatus of claim 4 , further comprising an input/output interface that provides the facial images and the PIN input from the key owner to the key generation model.
7. A face recognition based key generation method comprising:
learning a key generation model that is formed of a convolutional neural network (CNN) and a recurrent neural network (RNN) for outputting a consistent key value by using sample facial images of a key owner as inputs, learning the key generation model including:
outputting a feature value of the respective sample facial images, which are provided as inputs, from the CNN, and
performing iterative learning to output the consistent key using the feature value of each sample facial image and a personal identification number (PIN) as inputs, in the RNN; and
generating, using the key generation model, a key by receiving a facial image of the key owner at a desired key generation time,
wherein the key is for performing a digital signature.
8. The face recognition based key generation method of claim 7 , further comprising, when an error correcting code is included in the key output from the key generation model, correcting the output key value and eliminating the error correcting code.