Affective response predictor for a stream of stimuli

• US 8,886,581 B2
• Filed: 06/25/2011
• Issued: 11/11/2014
• Est. Priority Date: 05/11/2011
• Status: Active Grant

• Current Assignee: Affectomatics ltd.
• Original Assignee: Anat Thieberger Ben-Haim, Ari M. Frank, Gil Thieberger
• Inventors: Frank, Ari M., Thieberger, Gil, Thieberger Ben-Haim, Anat
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): FINK, THOMAS ANDREW
• Application Number:
  US13/168,971
• Publication Number:
  US 20120290520A1
• Time in Patent Office: 1,235 Days
• Field of Search: None
• US Class Current: 706/45
• 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

Predicting a user'"'"'s response to a stream of token instances, including: receiving a stream of token instances; partitioning the stream of token instances into consecutive temporal windows of token instances; predicting response of the user to temporal windows of token instances; predicting response of the user to a certain temporal window of token instances; and forwarding the prediction of the user to the stream of token instances.

Claims

1. A method for predicting affective response of a user to a stream of token instances, comprising:

receiving the stream of token instances;

partitioning the stream of token instances into consecutive temporal windows of token instances;

wherein the temporal windows of token instances are ordered according to their start time;

predicting affective response of the user to a first temporal window of token instances by providing a machine learning-based predictor with input data comprising;

a vector of values derived from the first temporal window of token instances, and an initial state value derived from a value of a measurement channel of the user taken at time corresponding to start of the first temporal window of token instances; and

for each successive temporal window of token instances after the first temporal window of token instances;

predicting affective response of the user to the successive temporal window of token instances by providing the machine learning-based predictor with input data comprising;

a vector of values derived from the successive temporal window of token instances, and an initial state value derived from prediction of affective response of the user to a temporal window of token instances preceding the successive temporal window of token instances.

2. The method of claim 1, further comprising forwarding a predicted affective response of the user to the stream of token instances;

wherein the predicted affective response comprises a value derived from a prediction of affective response of the user to a temporal window of token instances with a start time that occurs after the start time of the first temporal window of token instances.

3. The method of claim 1, wherein predictions of affective response of the user are expressed as at least one of:

values of a measurement channel of the user, and emotional responses.

4. The method of claim 1, wherein input data provided to the machine learning-based predictor further comprises information related to previous instantiations of token instances in order for the predictor to adjust prediction to account for habituation.

5. The method of claim 1, wherein input data provided to the machine learning-based predictor further comprises a baseline value for the user of a measurement channel of the user.

6. The method of claim 1, wherein input data provided to the machine learning-based predictor further comprises values corresponding to a situation of the user while the user is exposed to the token instances.

7. The method of claim 1, wherein the machine learning-based predictor utilizes a neural network.

8. The method of claim 1, wherein the machine learning-based predictor utilizes a regression model.

9. The method of claim 1, wherein the machine learning-based predictor utilizes a maximum entropy model classifier.

10. The method of claim 1, wherein the machine learning-based predictor utilizes a Hidden Markov Model.

11. A system configured to predict an affective response of a user to a stream of token instances, comprising:

a sample generator configured to receive the stream of token instances, and to partition the stream of token instances into consecutive temporal windows of token instances;

wherein the temporal windows of token instances are ordered according to their start time; and

a sequential machine learning-based predictor configured to receive input data comprising;

a vector of values derived from a first temporal window of token instances, and an initial state value derived from a value of a measurement channel of the user taken at time corresponding to start of the first temporal window of token instances; and

predict, based on the input data, an affective response of the user to the first temporal window of token instances; and

for each successive temporal window of token instances after the first temporal window of token instances, the sequential machine learning-based predictor is further configured to;

receive input data comprising;

a vector of values derived from the successive temporal window of token instances, and an initial state value derived from prediction of affective response of the user to a temporal window of token instances preceding the successive temporal window of token instances; and

to predict, based on the input data, an affective response of the user to the successive temporal window of token instances.

12. The system of claim 11, wherein the sequential machine learning-based predictor is further configured to forward a predicted affective response of the user to the stream of token instances;

wherein the predicted affective response comprises a value derived from a prediction of affective response of the user to a temporal window of token instances with a start time that occurs after the start time of the first temporal window of token instances.

13. The system of claim 11, wherein predictions of affective response of the user are expressed as at least one of:

values of a measurement channel of the user, and emotional responses.

14. The system of claim 11, wherein the input data provided to the sequential machine learning-based predictor further comprises information related to previous instantiations of token instances in order for the predictor to adjust prediction to account for habituation.

15. The system of claim 11, wherein the input data provided to the sequential machine learning-based predictor further comprises a baseline value for the user of a measurement channel of the user.

16. The system of claim 11, wherein the input data provided to the predictor further comprises values corresponding to a situation of the user while the user is exposed to the token instances.

17. The system of claim 11, wherein the sequential machine learning-based predictor utilizes a neural network.

18. The system of claim 11, wherein the sequential machine learning-based predictor utilizes a regression model.

19. The system of claim 11, wherein the sequential machine learning-based predictor utilizes a maximum entropy model classifier.

20. The system of claim 11, wherein the sequential machine learning-based predictor utilizes a Hidden Markov Model.