Methods for improving the performance of neural networks used for biometric authentication
A method of generating a biometric signature of a user for use in authentication using a neural network, the method comprising: receiving ( 110 ) a plurality of biometric samples from a user; extracting at least one feature vector using the plurality of biometric samples; using the elements of the at least one feature vector as inputs for a neural network; extracting the corresponding activations from an output layer of the neural network; and generating a biometric signature of the user using the extracted activations, such that a single biometric signature represents multiple biometric samples from the user.
1. A method of generating a biometric signature of a user for use in authentication using a neural network, the method comprising:
receiving a plurality of biometric samples from a user;
extracting at least one feature vector using the plurality of biometric samples;
using elements of the at least one feature vector as inputs for a neural network;
extracting corresponding activations from an output layer of the neural network; and
generating a biometric signature of the user using the extracted activations, such that a single biometric signature represents multiple biometric samples from the user;
wherein extracting the at least one feature vector using the plurality of biometric samples comprises extracting a feature vector from each of the plurality of biometric samples, resulting in a plurality of feature vectors; the steps of using the elements of the at least one feature vector as inputs for a neural network and extracting the corresponding activations from an output layer of the neural network comprise, for each feature vector from among the plurality of feature vectors, using the elements of said feature vector as inputs for the neural network and extracting the corresponding activations from an output layer of the neural network to generate a corresponding output vector, resulting in generation of a plurality of output vectors; and the step of generating the biometric signature of the user comprises taking a weighted average of the plurality of output vectors.
2. A method according to claim 1 , wherein after receiving the plurality of biometric samples from the user the method further comprises determining a signal to noise ratio of each of a plurality of biometric samples provided by the user and discarding one or more biometric samples with a signal to noise ratio below a predetermined threshold.
3. A method according to claim 2 , wherein after determining the signal to noise ratio of each of the plurality of biometric samples provided by the user the method further comprises discarding one or more biometric samples with a signal to noise ratio above a predetermined threshold.
4. A method according to claim 1 , wherein the biometric samples are vocal samples.
5. A method according to claim 4 , wherein after receiving the plurality of biometric samples from the user the method further comprises normalising the biometric samples.
6. A method according to claim 4 , wherein at least two of the plurality of biometric samples include environmental noise representative of different environments.
7. A method according to claim 4 , wherein after receiving the plurality of biometric samples from the user the method further comprises adding artificial noise to one or more of the plurality of biometric samples.
8. A method according to claim 4 , wherein extracting at least one feature vector using the plurality of biometric samples comprises:
concatenating two or more of the plurality of biometric samples and extracting a feature vector from the concatenated sample; or
concatenating all of the plurality of biometric samples and extracting a feature vector from the concatenated sample.
9. A method according to claim 1 , wherein weights of the weighted average are all equal.
10. A method according to claim 1 , wherein the method further comprises a step of prompting the user to provide further biometric samples.
11. A method according to claim 1 , wherein extracting a feature vector from a biometric sample comprises:
extracting low-level acoustic descriptors from said sample; or
extracting a feature vector from a biometric sample comprises using metadata associated with said sample.
12. A method of authenticating a user, the method comprising:
generating a biometric signature of a user, by:
receiving a plurality of biometric samples from a user;
extracting at least one feature vector using the plurality of biometric samples;
using elements of the at least one feature vector as inputs for a neural network;
extracting corresponding activations from an output layer of the neural network; and
generating a biometric signature of the user using the extracted activations, such that a single biometric signature represents multiple biometric samples from the user;
wherein extracting the at least one feature vector using the plurality of biometric samples comprises extracting a feature vector from each of the plurality of biometric samples, resulting in a plurality of feature vectors; the steps of using the elements of the at least one feature vector as inputs for a neural network and extracting the corresponding activations from an output layer of the neural network comprise, for each feature vector from among the plurality of feature vectors, using the elements of said feature vector as inputs for the neural network and extracting the corresponding activations from an output layer of the neural network to generate a corresponding output vector, resulting in generation of a plurality of output vectors; and the step of generating the biometric signature of the user comprises taking a weighted average of the plurality of output vectors;
receiving a biometric sample from the user;
extracting a feature vector from the biometric sample;
using the elements of the feature vector as inputs for a neural network;
extracting the corresponding activations from an output layer of the neural network to generate an output vector;
comparing said output vector with the biometric signature of the user; and
authorising the user based on the result of said comparison.