Methods and systems for detecting head motion during an authentication transaction

1. A method for detecting user head motion during an authentication transaction comprising:

generating, by a processor, a motion type feature vector and a user head motion type prediction based on data generated for a sequence of frames, the frames being included in biometric data captured from a user;

generating a liveness rating feature vector based on the generated frame data;

merging the motion type and liveness rating vectors;

generating a spoof prediction from the merged vector;

changing the user head motion type prediction to no motion when the generated spoof prediction indicates biometric data in the frames was spoofed;

storing the user head motion type prediction in a buffer, wherein the buffer stores predictions for different types of user head motion;

calculating a total number of predictions for each type of user head motion stored in the buffer;

calculating a score for each type of user head motion by dividing the total number of predictions for each type of user head motion by the total number of predictions in the buffer; and

determining the type of motion having the highest score as the final user head motion type detected for the frames.

2. The method for detecting user head motion in accordance with claim 1 , said generating a liveness rating feature vector step comprising:

generating a motion presence feature vector based on the generated frame data;

processing the motion presence feature vector to predict whether motion is present in the frames;

predicting whether the motion presence feature vector represents a spoof attack when motion is predicted to be present; and

storing the prediction in memory.

3. The method for detecting user head motion in accordance with claim 2 , said generating a motion presence feature vector step comprising:

calculating feature values from coordinate values, the coordinate values being included in the generated frame data and being from the last two frames of the sequence; and

generating the motion presence feature vector from the calculated feature values.

4. The method for detecting user head motion in accordance with claim 1 , said generating a motion type feature vector step comprising:

generating a signal based on the generated frame data, the signal having a regular processing rate that matches the sampling rate of a machine learning algorithm;

calculating feature values from a portion of the signal within a temporal window; and

generating the motion type feature vector from the calculated feature values.

5. The method for detecting user head motion in accordance with claim 4 , said generating a signal step comprising:

determining coordinate values within a temporal window, the coordinate values being included in the generated frame data;

dividing the temporal window into equal segments to define times;

calculating an interpolated coordinate value for each time based on the determined coordinate values; and

generating the signal from the interpolated coordinate values.

6. The method for detecting user head motion in accordance with claim 1 , further comprising communicating the final user head motion type and a corresponding score to a service provider website.

7. The method for detecting user head motion in accordance with claim 1 , said generating a user head motion type prediction step comprising processing the motion type feature vector with a machine learning algorithm.

8. A device for detecting user head motion during an authentication transaction comprising:

a processor; and

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

generating a motion type feature vector and a user head motion type prediction based on data generated for a sequence of frames, the frames being included in biometric data captured from a user;

generating a liveness rating feature vector based on the generated frame data;

merging the motion type and liveness rating vectors;

generating a spoof prediction from the merged vector;

changing the user head motion type prediction to no motion when the generated spoof prediction indicates biometric data in the frames was spoofed;

storing the user head motion type prediction in a buffer in said memory, wherein the buffer stores predictions for different types of user head motion;

determining a final user head motion type for the frames;

calculating a total number of predictions for each type of user head motion stored in the buffer;

calculating a score for each type of user head motion by dividing the total number of predictions for each type of user head motion by the total number of predictions in the buffer; and

determining the type of motion having the highest score as the final user head motion type detected for the frames.

9. The device for detecting user head motion in accordance with claim 8 , said generating a liveness rating feature vector step comprising:

generating a motion presence feature vector based on the generated frame data;

processing the motion presence feature vector to predict whether motion is present in the frames;

predicting whether the motion presence feature vector represents a spoof attack when motion is predicted to be present; and

storing the prediction in said memory.

10. The device for detecting user head motion in accordance with claim 9 , said generating a motion presence feature vector step comprising:

calculating feature values from coordinate values, the coordinate values being included in the generated frame data and being from the last two frames of the sequence; and

generating the motion presence feature vector from the calculated feature values.

11. The device for detecting user head motion in accordance with claim 8 , said generating a motion type feature vector step comprising:

generating a signal based on the generated frame data, the signal having a regular processing rate that matches the sampling rate of a machine learning algorithm;

calculating feature values from a portion of the signal within a temporal window; and

generating the motion type feature vector from the calculated feature values.

12. The device for detecting user head motion in accordance with claim 11 , said generating a signal step comprising:

determining coordinate values within a temporal window, the coordinate values being included in the generated frame data;

dividing the temporal window into equal segments to define times;

calculating an interpolated coordinate value for each time based on the determined coordinate values; and

generating the signal from the interpolated coordinate values.

13. The device for detecting user head motion in accordance with claim 8 , wherein the instructions further cause said device to perform steps comprising communicating the final user head motion type and a corresponding score to a service provider web site.

14. The device for detecting user head motion in accordance with claim 8 , wherein the instructions further cause said device to perform steps comprising processing the motion type feature vector with a machine learning algorithm.

15. A method for detecting user head motion during an authentication transaction comprising:

storing a head motion type prediction for a sequence of frames in a buffer, the buffer stores head motion type predictions for different types of motion, each head motion type prediction corresponding to the time a most recent frame in a sequence of frames was processed, the frames including biometric data captured from a user, wherein when biometric data for a sequence of frames is predicted to be spoofed an initial head motion type prediction for the sequence of frames is changed to no motion before said storing step;

calculating a total number of predictions for each type of user head motion stored in the buffer;

calculating a score for each type of user head motion by dividing the total number of predictions for each type of user head motion by the total number of predictions in the buffer; and

determining the type of motion having the highest score as the final user head motion type detected for the sequence of frames.

16. The method for detecting user head motion in accordance with claim 15 further comprising communicating the final user head motion type and the highest score to a service provider website.

17. The method for detecting user head motion in accordance with claim 15 , further comprising predicting the different user head motion types to be horizontal, vertical or no motion.