Biometric authentication through vascular studies
Introduced here are approaches to authenticating unknown persons based on variations in the spatial properties and directionality of blood flow through vessels over time. At a high level, these approaches rely on monitoring vascular dynamics to recognize unknown persons. For example, an authentication platform may examine digital images of an anatomical region to establish how a property of the vasculature within the anatomical region changed as a result of deformation. Examples of properties include the position, size, volume, and pressure of vessels included in the vasculature, as well as the velocity and acceleration of blood flowing through the vasculature.
1 . A method for producing a machine learning model that is able to predict blood flow through vasculature of an anatomical region, the method comprising:
receiving input that is indicative of a request to produce the machine learning model; and
in response to said receiving,
obtaining
(i) a first vascular pattern that provides spatial information regarding one or more vessels in the anatomical region while the anatomical region is in an undeformed state,
(ii) a second vascular pattern that provides spatial information regarding the one or more vessels while the anatomical region is in a deformed state,
(iii) a first flow pattern that indicates, for the first vascular pattern, how blood flows through the one or more vessels when the anatomical region is in the undeformed state, and
(iv) a second flow pattern that indicates, for the second vascular pattern, how blood flows through the one or more vessels when the anatomical region is in the deformed state;
providing the first vascular pattern, the second vascular pattern, the first flow pattern, and the second flow pattern to a machine learning algorithm that produces, as output, the machine learning model; and
storing the machine learning model in a database.
2 . The method of claim 1 ,
wherein the first vascular pattern is one of a first plurality of vascular patterns, each of which provides spatial information regarding the one or more vessels while the anatomical region is in the undeformed state,
wherein the second vascular pattern is one of a second plurality of vascular patterns, each of which provides spatial information regarding the one or more vessels while the anatomical region is in the deformed state.
3 . The method of claim 2 ,
wherein the first and second pluralities of vascular patterns are associated with an individual and acquired during a registration phase in which the individual is prompted to perform a plurality of instances of a gesture, and
wherein each of the plurality of instances is associated with a corresponding one of the first plurality of vascular patterns and a corresponding one of the second plurality of vascular patterns.
4 . The method of claim 2 , wherein the first and second pluralities of vascular patterns are associated with a plurality of individuals, each of whom is associated with at least one of the first plurality of vascular patterns and at least one of the second plurality of vascular patterns.
5 . The method of claim 1 , wherein the anatomical region is a finger.
6 . The method of claim 1 , wherein the anatomical region is a palmar side or a dorsal side of a hand.
7 . The method of claim 1 , wherein the anatomical region is a face.
8 . The method of claim 1 , wherein the machine learning model is a neural network with weights that are tuned as the machine learning algorithm learns from the first vascular pattern, the second vascular pattern, the first flow pattern, and the second flow pattern.