US 11,734,949 B1
System for biometric identification
Igor Kviatkovsky, Haifa (IL); Shunit Haviv, Haifa (IL); Manoj Aggarwal, Seattle, WA (US); Gal Novich, Ra'anana (IL); and Gerard Guy Medioni, Los Angeles, CA (US)
Assigned to AMAZON TECHNOLOGIES, INC., Seattle, WA (US)
Filed by AMAZON TECHNOLOGIES, INC., Seattle, WA (US)
Filed on Mar. 23, 2021, as Appl. No. 17/210,170.
Int. Cl. G06V 40/12 (2022.01); G06T 7/73 (2017.01); G06V 40/10 (2022.01); G06V 40/13 (2022.01)
CPC G06V 40/1365 (2022.01) [G06T 7/74 (2017.01); G06V 40/107 (2022.01); G06V 40/117 (2022.01); G06V 40/1318 (2022.01); G06T 2207/30196 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a camera;
a memory, storing first computer-executable instructions; and
a hardware processor to execute the first computer-executable instructions to:
acquire a first set of images of a hand using the camera;
determine a first canonical image depicting a palm of the hand and one or more fingers, wherein the first canonical image depicts the hand in a predetermined pose;
determine a first portion of the first canonical image that corresponds to the palm;
determine, based on the first portion, a first embedding vector;
determine a second portion of the first canonical image that corresponds to a first finger;
determine, based on the second portion, a second embedding vector;
determine a first candidate embedding vector associated with a first enrolled user identity, wherein the first candidate embedding vector is representative of a palm of a first enrolled user;
determine a second candidate embedding vector associated with a second enrolled user identity, wherein the second candidate embedding vector is representative of a palm of a second enrolled user;
determine a third candidate embedding vector associated with the first enrolled user identity, wherein the third candidate embedding vector is representative of a first finger of the first enrolled user;
determine a fourth candidate embedding vector associated with the second enrolled user identity, wherein the fourth candidate embedding vector is representative of a first finger of the second enrolled user;
determine a first distance in a first embedding space between the first embedding vector and the first candidate embedding vector;
determine a second distance in a second embedding space between the second embedding vector and the third candidate embedding vector;
determine a third distance by multiplying the first distance by the second distance;
determine a fourth distance in the first embedding space between the first embedding vector and the second candidate embedding vector;
determine a fifth distance in the second embedding space between the second embedding vector and the fourth candidate embedding vector;
determine a sixth distance by multiplying the fourth distance by the fifth distance; and
based on the third distance being less than the sixth distance, determine the first set of images are associated with the first enrolled user.