Methods and systems for determining user liveness

1. A method for determining user liveness comprising:

extracting, by a processor, overlapping differential signals from a first differential signal;

creating a principal component analysis coefficient set for each of the extracted overlapping differential signals;

selecting a principal component analysis coefficient subset from each set; and

generating an activity result for each of the extracted overlapping differential signals based on the respective subset.

2. A method for determining user liveness in accordance with claim 1 further comprising:

aggregating the activity results and comparing the aggregated activity results against a threshold score; and

determining the user is live when the aggregated activity result is equal to or greater than the threshold score.

3. A method for determining user liveness in accordance with claim 1 , said generating step comprising classifying the principal component analysis coefficient subset for each of the extracted overlapping differential signals using a machine learning technique.

4. A method for determining user liveness in accordance with claim 1 , said generating step comprising:

creating a vector for each subset;

inputting each vector into a support vector machines algorithm; and

generating the activity result for each of the extracted overlapping differential signals with the support vector machines algorithm.

5. A method for determining user liveness in accordance with claim 1 , said creating step comprising:

calculating spectral coefficients for each of the extracted overlapping differential signals;

normalizing the spectral coefficients;

projecting the normalized spectral coefficients onto a pre-trained principal component analysis basis; and

creating a principal component analysis coefficient set from the projected normalized spectral coefficients.

6. A method for determining user liveness in accordance with claim 1 , said selecting step comprising processing the principal component analysis subset of each of the extracted overlapping differential signals with a regressor to generate an activity result for each of the extracted overlapping differential signals.

7. A method for determining user liveness comprising:

creating, by a processor, a vector for each principal component analysis coefficient subset selected from a respective principal component analysis coefficient set, the coefficients created from captured user authentication data;

generating an activity result for each principal component subset and aggregating the activity results into a composite result; and

determining the user is live when the composite result is equal to or greater than a threshold score.

8. A method for determining user liveness in accordance with claim 7 , said creating step further comprising selecting the principal component analysis coefficients to include in each principal component subset to achieve maximum discrimination power.

9. A method for determining user liveness in accordance with claim 7 , further comprising:

capturing authentication data from the user;

calculating a differential surface area signal from the captured authentication data;

extracting a succession of overlapping sectional differential surface area signals from the calculated differential signal; and

creating a principal component analysis coefficient set for each of the extracted successive overlapping differential surface area signals.

10. A method for determining user liveness in accordance with claim 7 , further comprising:

extracting a succession of overlapping sectional differential surface area signals from a first differential signal;

calculating spectral coefficients from each extracted signal;

normalizing the spectral coefficients; and

projecting the normalized spectral coefficients onto a pre-trained principal component analysis basis to create the principal component analysis coefficient sets.

11. A method for determining user liveness in accordance with claim 7 , said generating step comprising:

generating a one when activity is detected in a subset; and

generating a zero when activity is not detected in a subset.

12. A method for determining user liveness in accordance with claim 7 further comprising:

calculating a mouth area for each frame in the captured user authentication data;

calculating a data quality value for each frame; and

assigning greater weight to mouth areas having higher quality values and lower weight to mouth areas having lower quality values.

13. A system for determining user liveness comprising:

a processor; and

a memory operable to store data, said processor being programmed to:

creating a vector for each principal component analysis coefficient subset selected from a respective principal component analysis coefficient set, the coefficients being created from captured user authentication data;

generate an activity result for each principal component subset and aggregating the activity results into a composite result; and

determine the user is live when the composite result is equal to or greater than a threshold score.

14. A system for determining user liveness in accordance with claim 13 , said processor being further programmed to select the principal component analysis coefficients to include in each principal component subset to achieve maximum discrimination power.

15. A system for determining user liveness in accordance with claim 13 , said processor being further programmed to:

calculate a differential surface area signal from captured authentication data;

extract a succession of overlapping sectional differential surface area signals from the calculated differential signal; and

create a principal component analysis coefficient set for each of the extracted succession of overlapping sectional differential surface area signals.

16. A system for determining user liveness in accordance with claim 13 , said processor being further programmed to:

extract a succession of overlapping sectional differential surface area signals from a first differential signal;

calculate spectral coefficients from each extracted signal;

normalize the spectral coefficients; and

project the normalized spectral coefficients onto a pre-trained principal component analysis basis to create the principal component analysis coefficient sets.

17. A system for determining user liveness in accordance with claim 13 , said processor being further programmed to:

calculate a mouth area for each frame in the captured user authentication data;

calculate a data quality value for each frame; and

assign greater weight to mouth areas having higher quality values and lower weight to mouth areas having lower quality values.