Methods and systems for determining user liveness

1. A method for determining user liveness comprising:

capturing, by a computing device, voice biometric data of a user during a verification transaction, the user and computing device being a user-computing device pair;

calculating a first matrix for the captured voice biometric data;

calculating an expansion coefficient matrix based on the first matrix and a record spectral shape matrix, the record spectral shape matrix being for the user-computing device pair;

calculating a distortion vector as the average of a ratio between the first matrix and the product of the record spectral shape matrix and the expansion coefficient matrix; calculating a spectral property difference as an element-wise function between the square value of the distortion vector and the square value of a user record normalization vector;

inputting the spectral property difference into a machine learning algorithm;

calculating an output score with the machine learning algorithm, the output score representing the difference between a probability the voice biometric data was captured from a live user and a probability the voice biometric data was not captured from a live user; and

determining the voice biometric data was captured from a live user when the output score satisfies a threshold score.

2. The method for determining user liveness in accordance with claim 1 further comprising determining the output score satisfies the threshold score when the output score is equal to or greater than the threshold score, the output score is greater than the threshold score, the output score is less than the threshold score, or the output score is equal to or less than the threshold score.

3. The method for determining user liveness in accordance with claim 1 further comprising calculating the expansion coefficient matrix using the equation Y≈HW, where Y is the first matrix, H is the expansion coefficient matrix, and W is the record spectral shape matrix.

4. The method for determining user liveness in accordance with claim 1 further comprising factorizing the first matrix with a non-negative factorization algorithm to estimate the spectral shape matrix.

5. The method for determining user liveness in accordance with claim 1 , said calculating a first matrix step comprising:

determining at least one portion of the captured voice biometric data;

calculating a spectral vector for the at least one portion; and

concatenating the spectral vectors to create the first matrix.

6. A computing device for determining user liveness comprising:

a processor; and

a memory configured to store data, said computing device being associated with a network and said memory being in communication with said processor and having instructions stored thereon which, when read and executed by said processor, cause said computing device to:

capture voice biometric data of a user during a verification transaction, the user and computing device being a user-computing device pair;

calculate a first matrix for the captured voice biometric data;

calculate an expansion coefficient matrix based on the first matrix and a record spectral shape matrix, the record spectral shape matrix being for the user-computing device pair;

calculate a distortion vector as the average of a ratio between the first matrix and the product of the record spectral shape matrix and the expansion coefficient matrix;

calculate a spectral property difference as an element-wise function between the square value of the distortion vector and the square value of a user record normalization vector;

input the spectral property difference into a machine learning algorithm;

calculate an output score with the machine learning algorithm, the output score representing the difference between a probability the voice biometric data was captured from a live user and a probability the voice biometric data was not captured from a live user; and

determine the voice biometric data was captured from a live user when the output score satisfies a threshold score.

7. The computing device for determining user liveness in accordance with claim 6 , wherein the instructions when read and executed by said processor, further cause said computing device to determine the output score satisfies the threshold score when the output score is equal to or greater than the threshold score, the output score is greater than the threshold score, the output score is less than the threshold score, or the output score is equal to or less than the threshold score.

8. The computing device for determining user liveness in accordance with claim 6 , wherein the instructions when read and executed by said processor, further cause said computing device to calculate the expansion coefficient matrix using the equation Y≈HW, where Y is the first matrix, H is the expansion coefficient matrix, and W is the record spectral shape matrix.

9. The computing device for determining user liveness in accordance with claim 6 , wherein the instructions when read and executed by said processor, further cause said computing device to factorize the first matrix with a non-negative factorization algorithm to estimate the spectral shape matrix.

10. The computing device for determining user liveness in accordance with claim 6 , wherein the instructions that cause said computing device to calculate a first matrix when read and executed by said processor, further cause said computing device to:

determine at least one portion of the captured voice biometric data;

calculate a spectral vector for the at least one portion; and

concatenate the spectral vectors to create the first matrix.

11. A computer program recorded on a non-transitory computer-readable recording medium included in a computing device for determining user liveness, the computer program being comprised of instructions, which when read and executed by the computing device cause the computing device to:

capture voice biometric data of a user during a verification transaction, the user and computing device being a user-computing device pair;

calculate a first matrix for the captured voice biometric data;

calculate an expansion coefficient matrix based on the first matrix and a record spectral shape matrix, the record spectral shape matrix being for the user-computing device pair;

calculate a distortion vector as the average of a ratio between the first matrix and the product of the record spectral shape matrix and the expansion coefficient matrix;

calculate a spectral property difference as an element-wise function between the square value of the distortion vector and the square value of a user record normalization vector;

input the spectral property difference into a machine learning algorithm;

calculate an output score with the machine learning algorithm, the output score representing the difference between a probability the voice biometric data was captured from a live user and a probability the voice biometric data was not captured from a live user; and

determine the voice biometric data was captured from a live user when the output score satisfies a threshold score.