Method and system for measuring shopper response to products based on behavior and facial expression

• US 8,219,438 B1
• Filed: 06/30/2008
• Issued: 07/10/2012
• Est. Priority Date: 06/30/2008
• Status: Active Grant

• Current Assignee: Videomining Corporation
• Original Assignee: Videomining Corporation
• Inventors: Sharma, Rajeev, Moon, Hankyu, Jung, Namsoon
• Primary Examiner(s): Choi, Peter
• Assistant Examiner(s): FEACHER, LORENA R
• Application Number:
  US12/215,879
• Time in Patent Office: 1,471 Days
• Field of Search: 705/7.29, 705/31, 705/33
• US Class Current: 705/7.29
• CPC Class Codes: G06F 18/295 Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities
• CPC Class Codes: G06Q 30/0204 Market segmentation
• CPC Class Codes: G06V 10/85 Markov-related models; Markov random fields
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: G06V 40/174 Facial expression recognition
• CPC Class Codes: G06V 40/193 Preprocessing; Feature extraction

Abstract

The present invention is a method and system for measuring human response to retail elements, based on the shopper'"'"'s facial expressions and behaviors. From a facial image sequence, the facial geometry—facial pose and facial feature positions—is estimated to facilitate the recognition of facial expressions, gaze, and demographic categories. The recognized facial expression is translated into an affective state of the shopper and the gaze is translated into the target and the level of interest of the shopper. The body image sequence is processed to identify the shopper'"'"'s interaction with a given retail element—such as a product, a brand, or a category. The dynamic changes of the affective state and the interest toward the retail element measured from facial image sequence is analyzed in the context of the recognized shopper'"'"'s interaction with the retail element and the demographic categories, to estimate both the shopper'"'"'s changes in attitude toward the retail element and the end response—such as a purchase decision or a product rating.

Claims

1. A method for determining a person'"'"'s response to a retail element, based on the person'"'"'s facial expression and shopping behavior, comprising the following steps of:

a) detecting and tracking a face from first input images captured by at least a first means for capturing images, estimating two-dimensional and three-dimensional poses of the face, and localizing facial features, using at least a control and processing system,

b) estimating gaze direction of the person using the two-dimensional and three-dimensional poses and positions of the facial features and changes in affective state of the person by extracting emotion-sensitive features, and recognizing a demographic category of the person,

c) detecting and tracking the person from second input images captured by at least a second means for capturing images, producing a trajectory of the person, and estimating body orientation, using the control and processing system,

d) identifying the shopping behaviors of the person toward the retail element, utilizing position and the body orientation of the person relative to the retail element, and

e) determining intermediate responses and end response of the person to the retail element by analyzing the changes in affective states and interest, in the context of the shopping behavior and the demographics category of the person,

wherein the first means for capturing images and the second means for capturing images are connected to the control and processing system via at least a means for video interface, and wherein the shopping behaviors include showing interest, engagement, interaction, or purchasing.

2. The method according to claim 1, wherein the method further comprises a step of training a plurality of first machines for localizing the face.

3. The method according to claim 1, wherein the method further comprises a step of training a plurality of second machines for localizing facial features in the face.

4. The method according to claim 1, wherein the method further comprises a step of extracting emotion-sensitive features from the face by applying emotion-sensitive feature filters to localized facial features and transient features.

5. The method according to claim 4, wherein the method further comprises a step of deriving emotion-sensitive feature filters, by first generating emotion-sensitive candidate filters and determining the emotion-sensitive feature filters by choosing the filters from the emotion-sensitive candidate filters that yield high responses to a sufficient number of facial images.

6. The method according to claim 5, wherein the method further comprises a step of generating the emotion-sensitive candidate filters by constructing a set of filters where each filter is designed to collect gradient response of a specific facial feature or a transient feature.

7. The method according to claim 1, wherein the method further comprises a step of determining facial muscle actions of the face based on the emotion-sensitive features.

8. The method according to claim 7, wherein the method further comprises a step of training a third machine so that the machine accepts the emotion-sensitive features as an input and computes a vector output where each component of the vector is the likelihood of the emotion-sensitive features arising from a specific facial muscle action.

9. The method according to claim 1, wherein the method further comprises a step of determining an affective state of the face by using a mapping from facial muscle actions to six emotional categories.

10. The method according to claim 1, wherein the method further comprises a step of determining an affective state of the face by using coordinates of six emotional categories in an affect space.

11. The method according to claim 1, wherein the method further comprises a step of determining the gaze direction of the person based on the three-dimensional facial pose and the positions of facial features.

12. The method according to claim 1, wherein the method further comprises a step of determining the body orientation of the person using a set of fourth learning machines wherein each machine is trained from images in a specific position in the view of the second means for capturing images.

13. The method according to claim 1, wherein the method further comprises a step of determining gross-level interest of the person based on the trajectory from tracking the person and the body orientation of the person.

14. The method according to claim 13, wherein the method further comprises a step of determining the gross-level interest of the person using a Bayesian graphical model wherein the orientations of the gross-level interest are modeled as hidden states.

15. The method according to claim 1, wherein the method further comprises a step of detecting the person'"'"'s engagement with the retail element based on a trajectory from tracking the person and the body orientation of the person.

16. The method according to claim 1, wherein the method further comprises a step of identifying the person'"'"'s interaction with the retail element based on a foreground object analysis between the person and a shelf.

17. The method according to claim 1, wherein the method further comprises a step of identifying the retail element with which the person is interacting based on a foreground object analysis and change detection on a shelf.

18. The method according to claim 1, wherein the method further comprises a step of detecting an incident of purchase and identifying the retail element that is purchased based on a foreground image analysis performed on a shelf and on a shopping cart or basket.

19. The method according to claim 1, wherein the method further comprises a step of segmenting the changes in affective states and interest based on the shopping behavior of the person,

wherein a sequence of the shopping behavior is divided into multiple segments, where each segment is a period in which the person shows a specific shopping behavior.

20. The method according to claim 1, wherein the method further comprises a step of determining the end response of the person toward the retail element using a Bayesian graphical model based on the changes in affective states and interest,

wherein the end response includes likelihood of purchase, a rating for a purchased product, and degree of satisfaction.

21. The method according to claim 20, wherein the method further comprises a step of decoding the intermediate responses of the person toward the retail element as the progression of hidden states in the Bayesian graphical model,

wherein the intermediate responses are defined as progression of the person'"'"'s attitude changes toward the retail element.

22. An apparatus for determining a person'"'"'s response to a retail element, based on the person'"'"'s facial expression and shopping behavior, comprising:

a) at least a first means for capturing images that captures first input images,

b) at least a control and processing system that performs the following steps of;

detecting and tracking a face from the first input images, estimating two-dimensional and three-dimensional pose of the face, localizing facial features,

estimating gaze direction of the person using the two-dimensional and three-dimensional poses and the positions of the facial features and changes in affective state of the person by extracting emotion-sensitive features,

c) at least a second means for capturing images that captures second input images,

d) the control and processing system that performs the following steps of;

detecting and tracking the person from the second input images producing a trajectory of the person, and estimating body orientation,

identifying the shopping behaviors of the person toward the retail element, utilizing position and the body orientation of the person relative to the retail element, and

determining intermediate responses and end response of the person to the retail element by analyzing the changes in affective states and interest, in the context of the shopping behavior, and

e) at least a means for video interface that connects the first means for capturing images and the second means for capturing images to the control and processing system,

wherein the shopping behaviors include showing interest, engagement, interaction, or purchasing.

23. The apparatus according to claim 22, wherein the apparatus further comprises means for training a plurality of first machines for localizing the face.

24. The apparatus according to claim 22, wherein the apparatus further comprises means for training a plurality of second machines for localizing facial features in the face.

25. The apparatus according to claim 22, wherein the apparatus further comprises means for extracting emotion-sensitive features from the face by applying emotion-sensitive feature filters to localized facial features and transient features.

26. The apparatus according to claim 25, wherein the apparatus further comprises means for deriving emotion-sensitive feature filters, by first generating emotion-sensitive candidate filters and determining the emotion-sensitive feature filters by choosing the filters from the emotion-sensitive candidate filters that yield high responses to a sufficient number of facial images.

27. The apparatus according to claim 26, wherein the apparatus further comprises means for generating the emotion-sensitive candidate filters by constructing a set of filters where each filter is designed to collect gradient response of a specific facial feature or a transient feature.

28. The apparatus according to claim 22, wherein the apparatus further comprises means for determining facial muscle actions of the face based on the emotion-sensitive features.

29. The apparatus according to claim 28, wherein the apparatus further comprises means for training a third machine so that the machine accepts the emotion-sensitive features as an input and computes a vector output where each component of the vector is the likelihood of the emotion-sensitive features arising from a specific facial muscle action.

30. The apparatus according to claim 22, wherein the apparatus further comprises means for determining an affective state of the face by using a mapping from facial muscle actions to six emotional categories.

31. The apparatus according to claim 22, wherein the apparatus further comprises means for determining an affective state of the face by using coordinates of six emotional categories in an affect space.

32. The apparatus according to claim 22, wherein the apparatus further comprises means for determining the gaze direction of the person based on the three-dimensional facial pose and the positions of facial features.

33. The apparatus according to claim 22, wherein the apparatus further comprises means for determining the body orientation of the person using a set of fourth learning machines wherein each machine is trained from images in a specific position in the view of the second means for capturing images.

34. The apparatus according to claim 22, wherein the apparatus further comprises means for determining gross-level interest of the person based on the trajectory from tracking the person and the body orientation of the person.

35. The apparatus according to claim 34, wherein the apparatus further comprises means for determining the gross-level interest of the person using a Bayesian graphical model wherein the orientations of the gross-level interest are modeled as hidden states.

36. The apparatus according to claim 22, wherein the apparatus further comprises means for detecting the person'"'"'s engagement with the retail element based on a trajectory from tracking the person and the body orientation of the person.

37. The apparatus according to claim 22, wherein the apparatus further comprises means for identifying the person'"'"'s interaction with the retail element based on a foreground object analysis between the person and a shelf.

38. The apparatus according to claim 22, wherein the apparatus further comprises means for identifying the retail element with which the person is interacting based on a foreground object analysis and change detection on a shelf.

39. The apparatus according to claim 22, wherein the apparatus further comprises means for detecting an incident of purchase and identifying the retail element that is purchased based on a foreground image analysis performed on a shelf and on a shopping cart or basket.

40. The apparatus according to claim 22, wherein the apparatus further comprises means for segmenting the changes in affective states and interest based on the shopping behavior of the person,

wherein a sequence of the shopping behavior is divided into multiple segments, where each segment is a period in which the person shows a specific shopping behavior.

41. The apparatus according to claim 22, wherein the apparatus further comprises means for determining the end response of the person toward the retail element using a Bayesian graphical model based on the changes in affective states and interest,

wherein the end response includes likelihood of purchase, a rating for a purchased product, and degree of satisfaction.

42. The apparatus according to claim 41, wherein the apparatus further comprises means for decoding the intermediate responses of the person toward the retail element as the progression of hidden states in the Bayesian graphical model,

wherein the intermediate responses are defined as progression of the person'"'"'s attitude changes toward the retail element.