Devices, methods and systems for biometric user recognition utilizing neural networks

• US 10,636,159 B2
• Filed: 03/13/2019
• Issued: 04/28/2020
• Est. Priority Date: 05/11/2015
• Status: Active Grant

• Current Assignee: Magic Leap, Inc.
• Original Assignee: Magic Leap, Inc.
• Inventors: Bradski, Gary R.
• Primary Examiner(s): Seth, Manev
• Application Number:
  US16/351,912
• Publication Number:
  US 20190213751A1
• Time in Patent Office: 412 Days
• CPC Class Codes: G06T 7/62 of area, perimeter, diameter or volume
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 40/15 Biometric patterns based on physiological signals, e.g. heartbeat, blood flow
• CPC Class Codes: G06V 40/18 Eye characteristics, e.g. of the iris
• CPC Class Codes: G06V 40/197 Matching; Classification

Abstract

A user identification system includes an image recognition network to analyze image data and generate shape data based on the image data. The system also includes a generalist network to analyze the shape data and generate general category data based on the shape data. The system further includes a specialist network to compare the general category data with a characteristic to generate narrow category data. Moreover, the system includes a classifier layer including a plurality of nodes to represent a classification decision based on the narrow category data.

Claims

1. A user identification system, comprising:

a processor configured to receive image data and run a neural network; and

a memory configured to store the image data and the neural network, the neural network comprising

an image recognition subnetwork configured to analyze image data and generate shape data based on the image data;

a generalist subnetwork configured to analyze the shape data and generate general category data based on the shape data;

a specialist layer configured to compare the general category data with a characteristic to generate narrow category data; and

a classifier layer including a plurality of nodes configured to represent a classification decision based on the narrow category data,

wherein the image recognition subnetwork is directly coupled to the generalist network,

wherein the generalist network is directly coupled to the specialist layer,

wherein the specialist layer is directly coupled to the classifier layer, and

wherein the neural network is a back propagation neural network including a plurality of layers.

2. The system of claim 1, wherein the back propagation neural network also includes error suppression and learning elevation.

3. The system of claim 1, further comprising an ASIC encoded with the image recognition network.

4. The system of claim 1, wherein the specialist layer comprises a back propagation network including a plurality of layers.

5. The system of claim 1, the neural network further comprising a tuning layer to modify the general category data based on user eye movements,

wherein the tuning layer is directly coupled to the generalist subnetwork and the specialist layer.

12. The system of claim 1, wherein the image recognition subnetwork, the generalist subnetwork, the specialist layer, and the classifier layer each comprise a plurality of nodes and a plurality of connectors, and

wherein each node of the plurality of nodes is operatively coupled to at least one other node of the plurality of nodes by a respective connector of the plurality of connectors.

13. The system of claim 12, wherein each connector of the plurality of connectors is bidirectional.

14. The system of claim 12, wherein the specialist layer comprises a plurality of nodes associated with a plurality of potentially confusing mismatched individuals in a database.

6. A method of identifying a user of a system, comprising:

analyzing image data;

generating shape data based on the image data;

analyzing the shape data;

generating general category data based on the shape data;

generating narrow category data by comparing the general category data with a characteristic;

generating a classification decision based on the narrow category data; and

generating a network of characteristics, wherein each respective characteristic of the network is associated with a potentially confusing mismatched individual in a database.

7. The method of claim 6, further comprising identifying an error in a piece of data.

8. The method of claim 7, further comprising suppressing the piece of data in which the error is identified.

9. The method of claim 6, wherein analyzing the image data comprises scanning a plurality of pixel of the image data.

10. The method of claim 6, wherein the image data corresponds to an eye of the user.

11. The method of claim 6, wherein the network of characteristics is generated when the system is first calibrated for the user.