Using electroencephalograph signals for task classification and activity recognition

• US 20070185697A1
• Filed: 02/07/2006
• Published: 08/09/2007
• Est. Priority Date: 02/07/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Tan, Desney, Lee, Johnny
• Granted Patent: US 7,580,742 B2
• Time in Patent Office: Days
• US Class Current: 703/11
• CPC Class Codes: A61B 5/16 Devices for psychotechnics using teaching or educational appliances G09B1/00 - G09B7/00; Testing reaction times ; Devices for evaluating the psychological state
• CPC Class Codes: A61B 5/164 Lie detection
• CPC Class Codes: A61B 5/369 Electroencephalography [EEG] devices for psychotechnics A61B5/16
• CPC Class Codes: A61B 5/372 Analysis of electroencephalograms
• CPC Class Codes: A61B 5/386 Accessories or supplementary instruments therefor
• CPC Class Codes: A61B 5/7207 of noise induced by motion artifacts
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G06F 2218/00 Aspects of pattern recognition specially adapted for signal processing
• CPC Class Codes: G06F 3/015 Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A method for classifying brain states in electroencephalograph (EEG) signals comprising building a classifier model and classifying brain states using the classifier model is described. Brain states are determined. Labeled EEG data is collected and divided into overlapping time windows. The time dimension is removed from each time window. Features are generated by computing the base features; combining the base features to form a larger feature set; pruning the large feature set; and further pruning the feature set for a particular machine learning technique. Brain states in unlabeled EEG data are classified using the classifier model by dividing the unlabeled EEG data into overlapping time windows and removing the time dimension from each time window. Features required by the classifier model are generated. Artifacts in the labeled and unlabeled EEG data comprise cognitive artifacts and non-cognitive artifacts.

Claims

1. A method for classifying brain states comprising building a classifier model using labeled electroencephalograph (EEG) data signals that include artifacts and classifying brain states in unlabeled EEG data signals that include artifacts using the classifier model.

2. The method of claim 1, wherein building a classifier model comprises:

a) determining brain states of interest;

b) collecting a set of labeled EEG data signals that include artifacts (“

labeled EEG data”

);

c) dividing the labeled EEG data into overlapping time windows;

d) removing the time dimension from each time window;

e) generating features from the time windows; and

f) building the classifier model.

3. The method of claim 2, wherein generating features from the time windows comprises;

a) computing the base features of the time windows;

b) combining the base features to form a larger feature set;

c) pruning the larger feature set without regard to machine learning techniques; and

d) further pruning the feature set for a machine learning technique.

4. The method of claim 1, wherein classifying brain states using the classifier model comprises:

a) collecting a set of unlabeled EEG data signals that include artifacts (“

unlabeled EEG data”

);

;

b) dividing the unlabeled EEG data into overlapping time windows;

c) removing the time dimension from each time window;

d) generating features required by the classifier model from the time windows;

e) using the classifier model to classify brain states in the unlabeled EEG data;

f) choosing the size of the averaging kernels; and

g) averaging the classifier output over adjacent time windows.

5. The method of claim 1, wherein the artifacts included in the labeled EEG data signals includes cognitive artifacts.

6. The method of claim 1, wherein the artifacts included in the labeled EEG data signals includes cognitive and non-cognitive artifacts.

7. The method of claim 1, wherein the artifacts included in the unlabeled EEG data signals includes cognitive artifacts.

8. The method of claim 1, wherein the artifacts included in the unlabeled EEG data signals includes cognitive and non-cognitive artifacts.

9. The method of claim 1, wherein the EEG signals are collected from pairs of sensors attached to individuals.

10. The method of claim 1, wherein the pairs of sensors comprise two pairs of sensors.

11. The method of claim 9, wherein individuals are neurologically healthy individuals.

12. The method of claim 11, wherein the neurologically healthy individuals are users in tests to evaluate user interfaces.

13. The method of claim 1, wherein the brain states are associated with at least one condition of users selected from the group of user conditions comprising:

interruptability, cognitive workload, task engagement, communication mediation, interpreting and predicting system response, surprise, satisfaction, and frustration.

14. The method of claim 1, including discriminating at least two different brain states.

15. The method of claim 1, including determining the persistence of brain states.

16. The method of claim 1, including determining the transitions between brain states.

17. The method of claim 1, wherein a plurality of features is generated and then pruned.