Vision testing via prediction-based setting of an initial stimuli characteristic for a user interface location

Patent metadata

Field	Value
Publication number	US-11039742-B1
Application number	US-202017083043-A
Country	US
Kind code	B1
Filing date	Oct 28, 2020
Priority date	Oct 28, 2020
Publication date	Jun 22, 2021
Grant date	Jun 22, 2021

Abstract

In some embodiments, initial feedback indicating threshold characteristics (under which a user sees initial stimuli presented on a user interface) may be provided to a prediction model, and a set of predicted characteristics (for a set of locations of the user interface) and a set of confidence scores associated with the set of locations may be obtained via the prediction model. Based on the set of confidence scores, one or more locations may be selected to be tested during a visual test presentation. As an example, the locations may be selected over one or more other locations of the set of locations based on the set of confidence scores. Based on predicted characteristics associated with the selected locations, stimuli may be presented at the selected locations during the visual test presentation. Visual defect information for the user may be generated based on feedback from the visual test presentation.

First claim

What is claimed is:

1. A system for facilitating vision testing via confidence-based selection of a testing location subset from a neural-network-provided set of locations, the system comprising:
a computer system that comprises one or more processors executing computer program instructions that, when executed, cause the computer system to perform operations comprising:
initializing, during a visual test presentation comprising multiple test rounds, a ground truth input for a neural network, the ground truth input comprising initial feedback related to an initial set of stimuli presented at an initial set of locations of a user interface of a wearable device under a range of characteristics, the initial feedback indicating threshold characteristics of the range of characteristics under which a user sees each stimulus of the initial set of stimuli;
performing the following during each round of the multiple test rounds:
providing, during the round of the visual test presentation, the ground truth input to the neural network to cause the neural network to generate (i) a set of predicted characteristics for a set of locations of the user interface and (ii) a set of confidence scores associated with the set of locations;
selecting, based on the set of confidence scores, locations of the set of locations that are to be tested during the round of the visual test presentation, the locations being selected over other locations of the set of locations based on confidence scores associated with the locations being greater than confidence scores associated with the other locations;
presenting, based on the selection of the locations and predicted characteristics associated with the selected locations, stimuli at the selected locations during the round without presenting stimuli at the other locations during the round such that:
(i) a first stimulus is initially presented at a first location of the locations based on a first predicted characteristic associated with the first location; and
(ii) a second stimulus is initially presented at second location of the locations based on a second predicted characteristic associated with the second location;

obtaining feedback related to the stimuli during the round, the feedback related to the stimuli indicating threshold characteristics of the range of characteristics under which the user sees each stimulus of the stimuli; and
updating, based on the feedback related to the stimuli, the ground truth input for a next round of the visual test presentation such that the updated ground truth input comprises the initial feedback and the feedback related to the stimuli; and

generating visual defect information for the user based on the initial feedback and the feedback from the multiple test rounds of the visual test presentation.

2. The system of claim 1 , wherein the first stimulus is initially presented at the first location under the first predicted characteristic.

3. The system of claim 1 , wherein the first stimulus is initially presented at the first location under at least one characteristic that is adjacent, in the range of characteristics, to the first predicted characteristic.

4. The system of claim 1 , wherein the range of characteristics comprises contrast levels, saturation levels, or sharpness levels.

5. A method comprising:
initializing a ground truth input for a prediction model, the ground truth input comprising initial feedback related to an initial set of stimuli presented at an initial set of locations of a user interface under a range of characteristics, the initial feedback indicating threshold characteristics of the range of characteristics under which a user sees each stimulus of the initial set of stimuli;
obtaining, via the prediction model, based on the ground truth input, (i) a set of predicted characteristics for a set of locations of the user interface and (ii) a set of confidence scores associated with the set of locations;
selecting, based on the set of confidence scores, one or more locations of the set of locations that are to be tested during a portion of a visual test presentation, the one or more locations being selected over one or more other locations of the set of locations based on the set of confidence scores;
presenting, based on one or more predicted characteristics associated with the selected locations, one or more stimuli at the selected locations during the portion of the visual test presentation;
obtaining feedback related to the one or more stimuli during the portion of the visual test presentation, the feedback indicating one or more threshold characteristics of the range of characteristics under which the user sees each stimulus of the one or more stimuli;
updating, based on the feedback related to the one or more stimuli, the ground truth input for a subsequent portion of the visual test presentation such that the updated ground truth input comprises the feedback related to the one or more stimuli; and
generating visual defect information for the user based on the initial feedback, the feedback from the portion of the visual test presentation, and additional feedback from the subsequent portion of the visual test presentation.

6. The method of claim 5 , wherein presenting the one or more stimuli at the selected locations comprises initially presenting a first stimulus at a first location of the selected locations based on a first predicted characteristic associated with the first location, and wherein the first predicted characteristic comprises a prediction of a threshold characteristic of the range of characteristics under which the user sees the first stimulus or a characteristic adjacent the threshold characteristic within the range of characteristics.

7. The method of claim 5 , wherein presenting the one or more stimuli at the selected locations comprises initially presenting a first stimulus at a first location of the selected locations under a first predicted characteristic associated with the first location, and
wherein the first predicted characteristic comprises a prediction of a threshold characteristic of the range of characteristics under which the user sees the first stimulus.

8. The method of claim 5 , wherein presenting the one or more stimuli at the selected locations comprises initially presenting a first stimulus at a first location of the selected locations under at least one characteristic that is adjacent, in the range of characteristics, to a first predicted characteristic associated with the first location, and
wherein the first predicted characteristic comprises a prediction of a threshold characteristic of the range of characteristics under which the user sees the first stimulus.

9. The method of claim 5 , wherein selecting the one or more locations comprises selecting the one or more locations over the one or more other locations based on one or more confidence scores associated with the one or more locations being greater than one or more confidence scores associated with the one or more other locations.

10. The method of claim 5 , wherein obtaining the set of predicted characteristics and the set of confidence scores comprises obtaining the set of predicted characteristics for the set of locations and the set of confidence associated with the set of locations from the prediction model subsequent to providing the ground truth input to the prediction model.

11. The method of claim 5 , further comprising:
obtaining, based on the updated ground truth input, (i) a second set of predicted characteristics for a second set of locations of the user interface and (ii) a second set of confidence scores associated with the second set of locations, the second set of locations comprising the o

Assignees

• Univ Miami

Inventors

• Abou Shousha Mohamed
• Elsawy Amr Saad Mohamed

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N20/00

  Machine learning · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title