Systems and methods for the detection of implicit gestures

• US 9,342,797 B2
• Filed: 04/03/2014
• Issued: 05/17/2016
• Est. Priority Date: 04/03/2014
• Status: Active Grant

• Current Assignee: Honda Motor Co., Ltd. (Honda Motor Company)
• Original Assignee: Edge Technologies, Inc, Honda Motor Co., Ltd. (Honda Motor Company)
• Inventors: El Dokor, Tarek A., King, Joshua, Cluster, Jordan, Holmes, James E.
• Primary Examiner(s): MARC-COLEMAN, MARTHE Y
• Application Number:
  US14/244,790
• Publication Number:
  US 20150286952A1
• Time in Patent Office: 775 Days
• Field of Search: 701/400, 701424-426, 701/537, 701/1, 345/156, 345/173, 455/41.2, 707/736, 707/758, 707/771, 709/203, 709/205, 709/201, 706/12
• US Class Current: 1/1
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Some embodiments provide systems and methods for enabling a learning implicit gesture control system for use by an occupant of a vehicle. The method includes identifying features received from a plurality of sensors and comparing the features to antecedent knowledge stored in memory. A system output action that corresponds to the features can then be provided in the form of a first vehicle output. The method further includes detecting a second vehicle output from the plurality of sensors and updating the antecedent knowledge to associate the system output action with the second vehicle output.

Claims

1. A method for enabling a learning implicit gesture control by an occupant of a vehicle, wherein an implicit gesture can include at least one input from the occupant of the vehicle, comprising:

identifying a first set of features in a plurality of inputs received from a plurality of sensors, wherein the first set of features are associated with a first vehicle output, the first vehicle output including activation of a feature of at least one of a plurality of system of the vehicle;

comparing the first set of features to an antecedent knowledge stored in a memory device, the antecedent knowledge including a plurality of previously extracted and stored features associated with a previously stored system output action, to identify a system output action corresponding to the first set of features and the first vehicle output, the system output action being configured to cause the first vehicle output;

detecting a second vehicle output from the plurality of inputs received from the plurality of sensors, wherein the second vehicle output is associated with an alternate vehicle output that is different than the first vehicle output; and

updating the antecedent knowledge to associate the system output action with the second vehicle output.

2. The method of claim 1, wherein updating the antecedent knowledge further includes updating the antecedent knowledge only upon detecting the second vehicle output occurs within a pre-determined time period of the system output action being identified.

3. The method of claim 1, wherein the antecedent knowledge includes knowledge stored from deep learning through a generalist network, a specialist network, and a monitor network.

4. The method of claim 1, wherein the plurality of sensors include visual sensors.

5. The method of claim 1, wherein the plurality of sensors include audio sensors.

6. The method of claim 1, wherein the system output action includes modifying a status of at least one of the plurality of systems of the vehicle.

7. The method of claim 6, wherein the plurality of systems of the vehicle include at least one of a reading light, a navigation system, and a telecommunications system.

8. A method for enabling learning implicit gesture control by an occupant of a vehicle based on at least one input from the occupant of the vehicle, comprising:

identifying a first set of features in a plurality of inputs received from a plurality of sensors, wherein the first set of features are associated with a first vehicle output;

querying the first set of features to a deep knowledge learning system stored in a memory device, the deep knowledge learning system including antecedent knowledge, to identify a system output action corresponding to the first set of features and the first vehicle output, the system output action being configured to cause the first vehicle output;

detecting a second vehicle output from the plurality of inputs received from the plurality of sensors, wherein the second vehicle output overrides the system output action begin configured to cause the first vehicle output; and

when the second vehicle output occurs within a pre-determined time period of the system output action being generated, performing a self-healing action to update the antecedent knowledge, wherein the self-healing action includes updating the antecedent knowledge to associate the system output action with the second vehicle output.

9. The method of claim 8, including, before updating the antecedent knowledge, prompting the occupant of the vehicle to confirm that the system output action should be associated with the second vehicle output.

10. The method of claim 8, wherein the self-healing action includes distinguishing the first set of features from a second set of features in the antecedent knowledge, the second set of features in the antecedent knowledge being associated with the system output action configured to cause the first vehicle output.

11. An implicit gesture and input learning system, comprising:

a plurality of systems of a vehicle and a plurality of sensors coupled to the vehicle;

a processor, the processor operatively connected for computer communication to the implicit gesture and input learning system, the vehicle, the plurality of system, and the plurality of sensors,

the processor being configured to;

receive a plurality of inputs from the plurality of sensors;

identify a first set of features in the plurality of inputs, wherein the first set of features are associated with a first vehicle output;

compare the first set of features to an antecedent knowledge stored in a memory device, the antecedent knowledge including a plurality of previously extracted and stored features, to identify a system output action corresponding to the first set of features and the first vehicle output, the system output action being configured to cause the first vehicle output;

detect a second vehicle output from the plurality of inputs received from the plurality of sensors, wherein the second vehicle output is a different vehicle output than the first vehicle output; and

update the antecedent knowledge to associate the system output action with the second vehicle output.

12. The system of claim 11, wherein the processor is configured to only update the antecedent knowledge when the second vehicle output occurs within a pre-determined time period of the system output action being generated.

13. The system of claim 11, wherein the plurality of previously extracted and stored features are associated with a given system output.

14. The system of claim 11, wherein the antecedent knowledge includes knowledge stored from deep learning through a generalist network, a specialist network, and a monitor network.

15. The system of claim 11, wherein the plurality of sensors include visual sensors.

16. The system of claim 11, wherein the plurality of sensors include audio sensors.

17. The system of claim 11, wherein the processor is further configured to execute the system output action thereby modifying a status of at least one of the plurality of systems of the vehicle.

18. The system of claim 17, wherein the plurality of systems of the vehicle include at least one of a reading light, a navigation system, and a telecommunications system.