Gaussian process-based approach for identifying correlation between wireless signals

Patent metadata

Field	Value
Publication number	US-9880257-B2
Application number	US-201514843961-A
Country	US
Kind code	B2
Filing date	Sep 2, 2015
Priority date	Sep 11, 2014
Publication date	Jan 30, 2018
Grant date	Jan 30, 2018

Abstract

Disclosed are apparatus and methods for providing outputs; e.g., location estimates, based on trained Gaussian processes modeling signals of wireless signal emitters. A computing device can determine first and second trained Gaussian processes. The respective first and second Gaussian processes can be based on first and second hyperparameter values related to first and second wireless signal emitters. The computing device can determine first and second sets of comparison hyperparameter values of the respective first and second hyperparameter values, and then determine whether the first and second sets of comparison hyperparameter values are within one or more threshold values. After determining that the first and second sets of comparison hyperparameter values are within the threshold(s), the computing device can determine the first and second Gaussian processes are dependent and then provide an estimated-location output based on a representative Gaussian process based on the first and the second Gaussian processes.

First claim

The invention claimed is:

1. A method, comprising:
determining, by a computing device, a plurality of trained Gaussian processes that model signals emitted by a plurality of wireless signal emitters, each Gaussian process of the plurality of trained Gaussian processes based on one or more hyperparameters, wherein the plurality of trained Gaussian processes comprises a first Gaussian process and a second Gaussian process, wherein the first Gaussian process is based on first hyperparameter values of the one or more hyperparameters related to a first wireless signal emitter of the plurality of wireless signal emitters, and wherein the second Gaussian process is based on second hyperparameter values of the one or more hyperparameters related to a second wireless signal emitter of the plurality of wireless signal emitters;
determining, by the computing device, a set of comparison hyperparameters from the one or more hyperparameters;
determining, by the computing device, a first set of comparison hyperparameter values of the first hyperparameter values and a second set of comparison hyperparameter values of the second hyperparameter values;
determining, by the computing device, whether the first set of comparison hyperparameter values are within one or more threshold values of the second set of comparison hyperparameter values;
after determining that the first set of comparison hyperparameter values are within the one or more threshold values of the second set of comparison hyperparameter values, determining by the computing device, that the first Gaussian process and the second Gaussian process are dependent Gaussian processes;
after determining that the first Gaussian process and the second Gaussian process are dependent Gaussian processes, determining by the computing device, a representative Gaussian process based on the first Gaussian process and the second Gaussian process; and
providing an estimated-location output of the computing device based on the representative Gaussian process.

2. The method of claim 1 , wherein providing the estimated-location output comprises:
receiving a request related to locating a mobile device;
determining an estimate of the location of the mobile device based on the representative Gaussian process;
generating the estimated-location output comprising the estimate of the location; and
providing the estimated-location output.

3. The method of claim 1 , further comprising:
after determining, for an outlying comparison hyperparameter of the set of comparison hyperparameters, that a first outlying comparison hyperparameter value of the first set of comparison hyperparameter values is not within a corresponding threshold outlying comparison hyperparameter value of a second outlying comparison hyperparameter value of the second set of hyperparameter values, determining by the computing device that the first Gaussian process is independent of the second Gaussian process.

4. The method of claim 1 , wherein determining by the computing device the representative Gaussian process comprises:
determining first signal-strength measurements used to train the first Gaussian process;
determining second signal-strength measurements used to train the second Gaussian process, wherein the first signal-strength measurements differ from the second signal-strength measurements; and
training the representative Gaussian process using both the first signal-strength measurements and the second signal-strength measurements.

5. The method of claim 4 , wherein providing the estimated-location output comprises:
after training the representative Gaussian process, providing the estimated-location output based on the representative Gaussian process.

6. The method of claim 1 , wherein the one or more hyperparameters are selected from the group of hyperparameters consisting of: a location hyperparameter, a power-output hyperparameter, a signal-attenuation hyperparameter, and a noise hyperparameter.

7. The method of claim 6 , wherein the location hyperparameter comprises a latitude hyperparameter and a longitude hyperparameter.

8. The method of claim 6 , wherein the noise hyperparameter comprises a background-noise hyperparameter and a noise-confidence hyperparameter.

9. The method of claim 1 , wherein the first wireless signal emitter is identified using a first wireless-signal-emitter identifier, wherein the second wireless signal emitter is identified using a second wireless-signal-emitter identifier, wherein the first Gaussian process is associated with the first wireless-signal-emitter identifier, and wherein the second Gaussian process is associated with the second wireless-signal-emitter identifier.

10. The method of claim 9 , wherein determining, by the computing device, the representative Gaussian process comprises:
determining first signal-strength measurements used to train the first Gaussian process;
determining second signal-strength measurements used to train the second Gaussian process, wherein the first signal-strength measurements differ from the second signal-strength measurements;
training the representative Gaussian process using both the first signal-strength measurements and the second signal-strength measurements; and
after training the representative Gaussian process, associating the representative Gaussian process with the first wireless-signal-emitter identifier and the second wireless-signal-emitter identifier.

11. The method of claim 9 , wherein at least one wireless-signal-emitter identifier of the first wireless-signal-emitter identifier and the second wireless-signal-emitter identifier comprises a Basic Service Set Identifier (BSSID).

12. A computing device, comprising:
one or more processors; and
data storage, configured to store at least computer-readable program instructions, wherein the instructions are configured to, upon execution by the one or more processors, cause the computing device to perform functions comprising:
determining a plurality of trained Gaussian processes that model signals emitted by a plurality of wireless signal emitters, each Gaussian process of the plurality of trained Gaussian processes based on one or more hyperparameters, wherein the plurality of trained Gaussian processes comprises a first Gaussian process and a second Gaussian process, wherein the first Gaussian process is based on first hyperparameter values of the one or more hyperparameters related to a first wireless signal emitter of the plurality of wireless signal emitters, and wherein the second Gaussian process is based on second hyperparameter values of the one or more hyperparameters related to a second wireless signal emitter of the plurality of wireless signal emitters;
determining a set of comparison hyperparameters from the one or more hyperparameters;
determining a first set of comparison hyperparameter values of the first hyperparameter values and a second set of comparison hyperparameter values of the second hyperparameter values;
determining whether the first set of comparison hyperparameter values are within one or more threshold values of the second set of comparison hyperparameter values;
after determining that the first set of comparison hyperparameter values are within the one or more threshold values of the second set of comparison hyperparameter values, determining that the first Gaussian process and the second Gaussian process are dependent Gaussian processes;
after determining that the first Gaussian process and the second Gaussian process are dependent Gaussian processes, determining a representative Gaussian process based on the first Gaussian process and the second Gaussian process; and
providing an estimated-lo

Assignees

• Google Inc
• Google Llc

Inventors

• Julian Brian John
• Williams Brian Patrick
• Le Grand Etienne

Classifications

• G01S5/0252

  Radio frequency fingerprinting · CPC title

• G01S5/0278 Primary

  involving statistical or probabilistic considerations (G01S5/0252, G01S5/0294 take precedence) · CPC title

• G01S5/0036

  of measured values, i.e. measurement on mobile and position calculation on base station · CPC title

• G01S5/02524

  Creating or updating the radio-map · CPC title