US 11,694,474 B2
Interactive user authentication
Pouria Mortazavian, London (GB); and Jacques Cali, London (GB)
Assigned to Onfido Ltd., London (GB)
Filed by ONFIDO LTD, London (GB)
Filed on Oct. 29, 2019, as Appl. No. 16/667,200.
Claims priority of application No. 18203125 (EP), filed on Oct. 29, 2018.
Prior Publication US 2020/0134148 A1, Apr. 30, 2020
Int. Cl. G06F 21/32 (2013.01); G06T 7/70 (2017.01); G06F 17/16 (2006.01); G06K 9/62 (2022.01); G06V 40/40 (2022.01)
CPC G06F 21/32 (2013.01) [G06F 17/16 (2013.01); G06K 9/6267 (2013.01); G06T 7/70 (2017.01); G06V 40/45 (2022.01)] 11 Claims
OG exemplary drawing
 
1. A computer-implemented method for verifying an electronic device user, the method comprising the steps of:
issuing at least one action instruction comprising a liveness challenge to an electronic device user using a notification mechanism of the electronic device;
recording response data from a plurality of data acquisition systems of the electronic device, the response data pertaining to the user's response to the at least one action instruction;
processing the response data to form classification scores, including identifying, from the response data, the likelihood of at least one characteristic pattern associated with an action instruction, comprising:
processing video data to assess a classification score of at least one characteristic motion associated with the action instruction by:
performing a plurality of head pose estimations on the video data,
processing the plurality of head pose estimations to form extracted pose information, including: extracting a series of angles from the plurality of head pose estimations; fitting a function to the series of angles; constructing a feature vector from parameters of the function, the fitting of the function, and the head pose estimations; and testing if the video data contains at least one characteristic motion with the feature vector, and
forming a facial action classification score using the extracted pose information;
combining the classification scores to form a classification value; and
verifying the user when the classification value satisfies a threshold.