Methods for predicting affective response from stimuli

• US 20120290514A1
• 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, Thieberger Ben-Haim, Anat
• Granted Patent: US 8,863,619 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 affective response predictor to predict a user'"'"'s emotional state after being exposed to tokens representing stimuli that influence the user'"'"'s affective state, comprising: receiving samples comprising temporal windows of token instances to which the user was exposed; 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; receiving target values, which represent affective response annotations of the user and correspond to the temporal windows of token instances; and creating the machine learning-based affective response predictor for the user, which compensates for non-linear effects resulting from the user being exposed to the subset of token instances originating from the same source and having overlapping instantiation periods, by running a machine learning training procedure on input data comprising the samples and the corresponding target values.

Claims

1. A method for training an emotional response predictor to predict a user'"'"'s emotional state after being exposed to tokens 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;

the token instances are spread over a long period of time, and a subset of the token instances originate from a same source and have overlapping instantiation periods;

receiving target values, which represent affective response annotations of the user and correspond to the temporal windows of token instances; and

training the emotional response predictor on input data comprising the samples and the corresponding target values;

wherein the emotional response predictor compensates for non-linear effects resulting from saturation due to the user being exposed to the subset of a plurality of token instances originating from the same source and having overlapping instantiation periods.

2. The method of claim 1, further comprising using emotional response 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 long period is over a week and spans different situations, and the resulting emotional response predictor captures situation dependent variations in the user'"'"'s affective response.

4. The method of claim 3, wherein training the emotional response predictor involves training at least two different machine learning-based emotional response predictors on data collected over periods where the user was in different situations.

5. The method of claim 1, wherein the training sets parameters of a regression model utilized by the predictor to make its predictions.

6. The method of claim 1, wherein the training sets weights of a neural network utilized by the predictor to make its predictions.

7. The method of claim 1, wherein the training sets parameters of a support vector machine for regression utilized by the predictor to make its predictions.

8. The method of claim 1, wherein the training sets parameters of a maximum entropy model classifier utilized by the predictor to make its predictions.

9. The method of claim 1, wherein the training sets parameters of a random forest model utilized by the predictor to make its predictions.

10. The method of claim 1, wherein the training sets parameters of a Naive Bayes classifier utilized by the predictor to make its predictions.

11. The method of claim 1, wherein the training comprises training multiple predictors;

and further comprising combining the predictions of the multiple predictors using an ensemble method.

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 user measurement channel values for the users state at the beginning of some of the temporal windows of token instances;

and generating the samples from the token instances and the initial user measurement channel values.

14. 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.

15. The method of claim 1, further comprising receiving additional samples comprising temporal windows of token instances that do not have corresponding target values, and the training of the emotional response predictor utilizes a semi-supervised training method.

16. A method for training a machine learning-based predictor to predict a users response to token instances representing stimuli that influence the users response, the method comprising:

receiving samples comprising temporal windows of token instances comprising representations of elements extracted from digital media content;

the token instances exposed to the user are spread over a long period of time, and a subset of the token instances originate from a same digital media source and have overlapping instantiation periods;

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 predictor to predict values representing the user'"'"'s response after being exposed to token instances;

the machine learning-based predictor compensates for non-linear effects resulting from the user being exposed to the subset of token instances originating from the same digital media source and having overlapping instantiation periods, by running a machine learning training procedure on input data comprising the samples and the corresponding target values.

17. The method of claim 16, wherein the target values are affective response annotations, and the machine learning-based predictor is a machine learning-based affective response predictor.

18. The method of claim 16, wherein the target values are derived from user measurement channels of the user, and the machine learning-based predictor expresses the expected response to tokens in terms of the user measurement channels'"'"' values.

19. The method of claim 16, 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.

20. A method for training a machine learning-based predictor to predict a user'"'"'s response measurement channel value after being exposed to tokens representing stimuli that influence the user'"'"'s affective state that is expressed through values of the measurement channel, the method comprising:

receiving samples comprising temporal windows of token instances to which the user was exposed;

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;

receiving target values, which represent the value of the user'"'"'s measurement channel and correspond to the temporal windows of token instances; and

training the machine learning-based predictor on input data comprising the samples and the corresponding target values to predict values of the user'"'"'s measurement channel after being exposed to token instances;

wherein the machine learning-based predictor compensates for non-linear effects resulting from the user being exposed to the subset of token instances originating from the same source and having overlapping instantiation periods.