Habituation-compensated predictor of affective response

• US 20120290516A1
• Filed: 06/25/2011
• Published: 11/15/2012
• Est. Priority Date: 05/11/2011
• Status: Active Grant

• Current Assignee: Affectomatics ltd.
• Original Assignee: Affectivon Limited
• Inventors: Frank, Ari M., Thieberger, Gil, Ben-Haim, Anat Thieberger
• Granted Patent: US 9,076,108 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: G06F 3/013 Eye tracking input arrangements G06F3/015 takes precedence
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation

Abstract

Creating a machine learning-based habituation-compensated predictor of a user'"'"'s response to token instances representing stimuli that influence the user'"'"'s affective state, comprising: receiving samples comprising temporal windows of token instances to which the user was exposed, wherein the token instances have overlapping instantiation periods; the samples further comprise data on previous instantiations of at least one of the token instances from the temporal windows; receiving target values corresponding to the temporal windows of token instances; the target values represent the user'"'"'s responses to the token instances from the temporal windows of token instances; training the machine learning-based habituation-compensated predictor to predict the user'"'"'s response to token instances, while accounting for the influence of the user'"'"'s previous exposure to tokens; wherein the training uses the samples, the data on previous instantiations, and the corresponding target values

Claims

1. A method for training a machine learning-based habituation-compensated predictor of a user'"'"'s response to token instances representing stimuli that influence the user'"'"'s affective state, the method comprising:

receiving samples comprising temporal windows of token instances to which the user was exposed, wherein the token instances have overlapping instantiation periods;

the samples further comprise data on previous instantiations of at least one of the token instances from the temporal windows;

receiving target values corresponding to the temporal windows of token instances;

the target values represent the user'"'"'s responses to the token instances from the temporal windows of token instances;

utilizing the samples, the data on previous instantiations, and the corresponding target values for training the machine learning-based habituation-compensated predictor to predict the user'"'"'s response to token instances, while accounting for the influence of the user'"'"'s previous exposure to tokens.

2. The method of claim 1, further comprising using the machine learning-based habituation-compensated predictor for predicting the user'"'"'s emotional state after being exposed to a sample comprising a new temporal window of token instances.

3. The method of claim 1, wherein the target values are affective response annotations, and the machine learning-based habituation-compensated predictor predicts the user'"'"'s expected affective response to token instances.

4. The method of claim 1, wherein the target values are derived from user measurement channels of the user, and the machine learning-based habituation-compensated predictor for the user expresses the expected response to tokens in terms of the user measurement channels'"'"' values.

5. The method of claim 1, wherein the data on previous instantiations comprises the time elapsed since the last time the user was exposed to a similar token instance.

6. The method of claim 5, further comprising a classifier configured to classify token instances into a predefined set of classes;

and similar token instances refer to token instances belonging to the same class.

7. The method of claim 1, wherein the machine learning-based habituation-compensated predictor is generated by analyzing at least two different machine learning-based user response models that were trained on data collected over periods where the user was in different situations;

and the habituation-compensated predictor predicts essentially differently the user'"'"'s expected responses to essentially the same tokens in the different situations.

8. The method of claim 1, wherein the step of training the machine learning-based habituation-compensated predictor comprises training a neural network.

9. The method of claim 1, wherein the step of training the machine learning-based habituation-compensated predictor comprises training a regression model.

10. The method of claim 1, wherein the step of training the machine learning-based habituation-compensated predictor comprises training a maximum entropy model classifier.

11. The method of claim 1, wherein the token instances are spread over a long period of time, and a subset of the token instances originate from same source and have overlapping instantiation periods.

12. The method of claim 1, further comprising receiving the token instances and a baseline value for the user;

and generating the samples from the token instances and the baseline value for the user.

13. The method of claim 1, further comprising receiving the token instances and initial affective response values for the user'"'"'s state at the beginning of some of the temporal windows of token instances;

and generating the samples from the token instances and the affective response value for the user.

14. A machine learning-based habituation-compensated predictor of a user'"'"'s response to token instances representing stimuli that influence the user'"'"'s affective state, comprising:

a machine learning trainer configured to receive samples comprising temporal windows of token instances to which the user was exposed, wherein the token instances have overlapping instantiation periods;

the samples further comprise data on previous instantiations of at least one of the token instances from the temporal windows;

the machine learning trainer is further configured to receive target values corresponding to the temporal windows of token instances;

the target values represent the user'"'"'s responses to the token instances from the temporal windows of token instances;

the machine learning trainer is further configured to utilize the samples, the data on previous instantiations, and the corresponding target values to train the machine learning-based habituation-compensated predictor to predict the user'"'"'s response to token instances, while accounting for the influence of the user'"'"'s previous exposure to tokens.

15. The machine of claim 14, further comprising a sample generator configured to receive the token instances and the data on previous instantiations, and configured to provide the samples.

16. The machine of claim 14, wherein the target values are affective response annotations, and the machine learning-based habituation-compensated predictor predicts the user'"'"'s expected affective response to token instances.

17. The machine of claim 16, further comprising a target value generator configured to receive user measurement channels of the user, and configured to provide the affective response annotations.

18. The machine of claim 14, wherein the target values are derived from user measurement channels of the user, and the machine learning-based habituation-compensated predictor for the user expresses the expected response to tokens in terms of the user measurement channels'"'"' values.

19. The machine of claim 14, wherein the data on previous instantiations comprises the time elapsed since the last time the user was exposed to a similar token instance;

and further comprising a classifier component configured to classify token instances into a predefined set of classes; and

similar token instances refer to token instances belonging to the same class.

20. The machine of claim 14, wherein the token instances are spread over a long period of time, and a subset of the token instances originate from same source and have overlapping instantiation periods.