Granted Patent B2

US 12,265,167 · App. 17/930,644 · Granted Apr 1, 2025

Inventors: Tianwei Xing (Santa Clara, CA); Wenjun Jiang (San Jose, CA); Xun Chen (Fremont, CA)

Assignee: Samsung Electronics Co., Ltd.

G01S5/02521H04W64/003

Abstract

Claims (67)

1. A method comprising:

receiving wireless fingerprint data associated with a specified area, the wireless fingerprint data identifying (i) multiple locations within the specified area and (ii) for each location, one or more signal strength values associated with wireless signals received at the location from one or more of multiple wireless transmitters, wherein the wireless fingerprint data is missing signal strength values for one or more of the wireless transmitters at one or more specific locations within the specified area;

generating a first training dataset by adding filler signal strength values in place of at least some of the signal strength values that are missing from the wireless fingerprint data;

training a first machine learning model using the first training dataset, the first machine learning model trained to receive a specified location as input and generate predicted signal strength values as outputs; and

using the trained first machine learning model to generate additional signal strength values, at least some of the additional signal strength values to be used in place of at least a portion of the signal strength values that are missing from the wireless fingerprint data.

2. The method of claim 1 , wherein:

the trained first machine learning model is used to generate a second training dataset, the second training dataset including the additional signal strength values generated using the trained first machine learning model; and

the method further comprises training a second machine learning model using the second training dataset, the second machine learning model trained to receive input signal strength values and generate a predicted location based on the input signal strength values.

3. The method of claim 2 , further comprising:

deploying the trained second machine learning model to a mobile electronic device for use in predicting a location of the mobile electronic device based on wireless signals received by the mobile electronic device.

4. The method of claim 2 , wherein:

the first machine learning model comprises a neural network; and

the second machine learning model comprises a weighed K-nearest neighbor model.

5. The method of claim 1 , wherein training the first machine learning model comprises minimizing an error between (i) the signal strength values and the filler signal strength values contained in the first training dataset and (ii) predicted signal strength values generated by the first machine learning model during the training.

6. The method of claim 1 , wherein training the first machine learning model comprises validating performance of the first machine learning model using a masked loss that is based on at least some of the signal strength values contained in the wireless fingerprint data.

7. The method of claim 1 , wherein each filler signal strength value is (i) associated with location coordinates within the specified area and (ii) determined using one or more signal strength values that are within a threshold distance of the location coordinates.

8. An apparatus comprising:

at least one processing device configured to:

receive wireless fingerprint data associated with a specified area, the wireless fingerprint data identifying (i) multiple locations within the specified area and (ii) for each location, one or more signal strength values associated with wireless signals received at the location from one or more of multiple wireless transmitters, wherein the wireless fingerprint data is missing signal strength values for one or more of the wireless transmitters at one or more specific locations within the specified area;

generate a first training dataset by adding filler signal strength values in place of at least some of the signal strength values that are missing from the wireless fingerprint data;

train a first machine learning model using the first training dataset, the first machine learning model trained to receive a specified location as input and generate predicted signal strength values as outputs; and

use the trained first machine learning model to generate additional signal strength values, at least some of the additional signal strength values to be used in place of at least a portion of the signal strength values that are missing from the wireless fingerprint data.

9. The apparatus of claim 8 , wherein:

the at least one processing device is configured to use the trained first machine learning model to generate a second training dataset, the second training dataset including the additional signal strength values generated using the trained first machine learning model; and

the at least one processing device is further configured to train a second machine learning model using the second training dataset, the second machine learning model trained to receive input signal strength values and generate a predicted location based on the input signal strength values.

10. The apparatus of claim 9 , wherein the at least one processing device is further configured to deploy the trained second machine learning model to a mobile electronic device for use in predicting a location of the mobile electronic device based on wireless signals received by the mobile electronic device.

11. The apparatus of claim 9 , wherein:

the first machine learning model comprises a neural network; and

the second machine learning model comprises a weighed K-nearest neighbor model.

12. The apparatus of claim 8 , wherein, to train the first machine learning model, the at least one processing device is configured to minimize an error between (i) the signal strength values and the filler signal strength values contained in the first training dataset and (ii) predicted signal strength values generated by the first machine learning model during the training.

13. The apparatus of claim 8 , wherein, to train the first machine learning model, the at least one processing device is configured to validate performance of the first machine learning model using a masked loss that is based on at least some of the signal strength values contained in the wireless fingerprint data.

14. The apparatus of claim 8 , wherein each filler signal strength value is (i) associated with location coordinates within the specified area and (ii) determined using one or more signal strength values that are within a threshold distance of the location coordinates.

15. A non-transitory computer readable medium containing instructions that when executed cause at least one processor to:

receive wireless fingerprint data associated with a specified area, the wireless fingerprint data identifying (i) multiple locations within the specified area and (ii) for each location, one or more signal strength values associated with wireless signals received at the location from one or more of multiple wireless transmitters, wherein the wireless fingerprint data is missing signal strength values for one or more of the wireless transmitters at one or more specific locations within the specified area;

generate a first training dataset by adding filler signal strength values in place of at least some of the signal strength values that are missing from the wireless fingerprint data;

train a first machine learning model using the first training dataset, the first machine learning model trained to receive a specified location as input and generate predicted signal strength values as outputs; and

use the trained first machine learning model to generate additional signal strength values, at least some of the additional signal strength values to be used in place of at least a portion of the signal strength values that are missing from the wireless fingerprint data.

16. The non-transitory computer readable medium of claim 15 , wherein:

the instructions when executed cause the at least one processor to use the trained first machine learning model to generate a second training dataset, the second training dataset including the additional signal strength values generated using the trained first machine learning model; and

the non-transitory computer readable medium further contains instructions that when executed cause the at least one processor to train a second machine learning model using the second training dataset, the second machine learning model trained to receive input signal strength values and generate a predicted location based on the input signal strength values.

17. The non-transitory computer readable medium of claim 16 , wherein the non-transitory computer readable medium further contains instructions that when executed cause the at least one processor to deploy the trained second machine learning model to a mobile electronic device for use in predicting a location of the mobile electronic device based on wireless signals received by the mobile electronic device.

18. The non-transitory computer readable medium of claim 16 , wherein:

the first machine learning model comprises a neural network; and

the second machine learning model comprises a weighed K-nearest neighbor model.

19. The non-transitory computer readable medium of claim 15 , wherein the instructions that when executed cause the at least one processor to train the first machine learning model comprise:

instructions that when executed cause the at least one processor to minimize an error between (i) the signal strength values and the filler signal strength values contained in the first training dataset and (ii) predicted signal strength values generated by the first machine learning model during the training.

20. The non-transitory computer readable medium of claim 15 , wherein the instructions that when executed cause the at least one processor to train the first machine learning model comprise:

instructions that when executed cause the at least one processor to validate performance of the first machine learning model using a masked loss that is based on at least some of the signal strength values contained in the wireless fingerprint data.

21. A method comprising:

obtaining, by a mobile electronic device, signal strength values associated with wireless signals received at the mobile electronic device;

providing the obtained signal strength values as input to a trained location prediction machine learning model;

using the trained location prediction machine learning model to generate a predicted location of the mobile electronic device; and

presenting the predicted location of the mobile electronic device on a display of the mobile electronic device;

wherein the trained location prediction machine learning model is trained using (i) wireless fingerprint data collected within a specified area and (ii) imputed or augmented signal strength values generated using a trained generator machine learning model.

22. The method of claim 21 , wherein the trained location prediction machine learning model is trained by:

obtaining the wireless fingerprint data collected within the specified area, the wireless fingerprint data identifying (i) multiple locations within the specified area and (ii) for each location, one or more signal strength values associated with wireless signals received at the location from one or more of multiple wireless transmitters, wherein the wireless fingerprint data is missing signal strength values for one or more of the wireless transmitters at one or more specific locations within the specified area;

generating a first training dataset by adding filler signal strength values in place of at least some of the signal strength values that are missing from the wireless fingerprint data;

training a generator machine learning model using the first training dataset, the generator machine learning model trained to receive a specified location as input and generate predicted signal strength values as outputs;

using the trained generator machine learning model to generate a second training dataset, the second training dataset including the imputed or augmented signal strength values produced using the trained generator machine learning model, at least some of the imputed or augmented signal strength values used in place of at least a portion of the signal strength values that are missing from the wireless fingerprint data; and

training a location prediction machine learning model using the second training dataset to produce the trained location prediction machine learning model.

23. The method of claim 21 , wherein the trained location prediction machine learning model comprises a weighed K-nearest neighbor model, the weighed K-nearest neighbor model configured to:

identify K wireless fingerprints from training data that are closest or most similar to a wireless fingerprint formed using the obtained signal strength values; and

perform a weighted combination of known locations associated with the K wireless fingerprints in order to generate the predicted location of the mobile electronic device.

24. The method of claim 23 , wherein the K wireless fingerprints are identified based on one of: Euclidean distances, cosine similarities, and correlation coefficients.

25. The method of claim 23 , wherein the weighted combination is determined using one of: a uniform weighting, a reciprocal weighting, a Gaussian kernel weighting, and an Epanechnikov kernel weighting.

26. The method of claim 21 , further comprising:

receiving the trained location prediction machine learning model from a server that deploys the trained location prediction machine learning model to the mobile electronic device.

Assignments (1)

Continuity (2)

References Cited (24)