Detecting events from features derived from multiple ingested signals

• US 10,474,733 B2
• Filed: 04/09/2019
• Issued: 11/12/2019
• Est. Priority Date: 02/09/2018
• Status: Expired due to Fees

• Current Assignee: Banjo, Inc.
• Original Assignee: Banjo, Inc.
• Inventors: Patton, Damien, Mehta, Rish, Bruckhaus, Tilmann
• Primary Examiner(s): Aspinwall, Evan
• Application Number:
  US16/379,401
• Publication Number:
  US 20190251139A1
• Time in Patent Office: 217 Days
• Field of Search: 707737
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/29 Geographical information databases
• CPC Class Codes: G06F 16/9537 Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 18/20 Analysing
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks
• CPC Class Codes: H04W 4/90 Services for handling of emergency or hazardous situations, e.g. earthquake and tsunami warning systems [ETWS]

Abstract

The present invention extends to methods, systems, and computer program products for detecting events from features derived from multiple signals. In one aspect, an event detection infrastructure determines that characteristics of multiple signals, when considered collectively, indicate an event of interest to one or more parties. In another aspect, an evaluation module determines that characteristics of one or more signals indicate a possible event of interest to one or more parties. A validator then determines that characteristics of one or more other signals validate the possible event as an actual event of interest to the one or more parties. Signal features can be used to compute probabilities of events occurring.

Claims

1. A method implemented at a computing system comprising a processor:

receiving a first normalized signal including a first location dimension and a first context dimension from a first hardware computing component associated with a signal source, the first context dimension including a first single source probability representing at least a first approximate probability of a real-world event of a specified event type;

deriving first one or more features from the first normalized signal including from the first single source probability;

determining that the first one or more features, including the first single source probability, are below a first threshold to be identified on their own as the real-world event of the specified event type;

receiving a second normalized signal including a second location dimension and a second context dimension from a second hardware computing component associated with a second signal source, the second context dimension including a second single source probability representing at least a second approximate probability of the real-world event of the specified event type;

calculating a multisource probability by at least aggregating the first single source probability that is below the first threshold to be identified as the real-world event and the second single source probability; and

detecting the real-world event from evidence provided by the multisource probability, including determining that the multisource probability exceeds a second threshold probability associated with the specified event type.

2. The method of claim 1, wherein determining that the first one or more features, including the first single source probability, are below the first threshold to be identified as the real-world event comprises detecting a possible event from the first one or more features;

and

wherein detecting the real-world event from evidence provided by the multisource probability comprises validating the possible event as the real-world event based on the second one or more features.

3. The method of claim 1, further comprising:

including the first normalized signal in a signal sequence;

determining that the second normalized signal meets a threshold temporal similarity to the first normalized signal;

determining that the second normalized signal meets a threshold spatial similarity to the first normalized signal; and

including the second normalized signal in the signal sequence.

4. The method of claim 3, wherein aggregating the first single source probability with the second single source probability comprises deriving features of the signal sequence.

5. The method of claim 4, wherein deriving features of the signal sequence comprises deriving one or more of:

a percentage, a count, a histogram, or a duration.

6. The method of claim 3, wherein determining that the second normalized signal meets the threshold spatial similarity to the first normalized signal comprises determining that the first location dimension and the second location dimension are within a specified distance of one another.

7. The method of claim 3, wherein determining that the second normalized signal meets the threshold spatial similarity to the first normalized signal comprises determining that the first location dimension and the second location dimension are within a specified number of geo cells of one another.

8. The method of claim 3, further comprising determining that the second normalized signal is not a duplicate of the first normalized signal prior to including the second normalized signal in the signal sequence.

9. The method of claim 1, wherein the first normalized signal corresponds to one of:

a social post with geographic content, a social post without geographic content, an image from a camera feed, a 911 call, weather data, IoT device data, satellite data, satellite imagery, a sound clip from a listening device, data from air quality sensors, a sound clip from radio communication, crowd sourced traffic information, or crowd sourced road information.

10. The method of claim 9, wherein the second normalized signal corresponds to a different one of:

a social post with geographic content, a social post without geographic content, an image from a traffic camera feed, a 911 call, weather data, IoT device data, satellite data, satellite imagery, a sound clip from a listening device, data from air quality sensors, a sound clip from radio communication, crowd sourced traffic information, or crowd sourced road information.

11. A system comprising:

a processor;

system memory coupled to the processor and storing instructions configured to cause the processor to;

receive a first normalized signal including a first location dimension and a first context dimension, the first context dimension including a first single source probability representing at least a first approximate probability of a real-world event of a specified event type;

derive first one or more features from the first normalized signal including from the first single source probability;

determine that the first one or more features, including the first single source probability, does not meet a first threshold level of evidence to be identified on their own as the real-world event of the specified event type;

receive a second normalized signal including a second location dimension and a second context dimension from a second hardware computing component associated with a second signal source, the second context dimension including a second single source probability representing at least a second approximate probability of the real-world event of the specified event type;

determine a multisource probability comprising an aggregate of the first single source probability that is below the first threshold to be identified as the real-world event and the second single source probability; and

detect the real-world event from evidence provided by the multisource probability, including a determination that the multisource probability exceeds a second threshold probability associated with the event type.

12. The system of claim 11, wherein instructions configured to determine that the first one or more features, including the first single source probability, does not meet the threshold level of evidence to be identified as the real-world event comprises instructions configured to determine the first single source probability is not confirming of the real-world event but does meet a different threshold level of evidence to warrant further investigation of the event type.

13. The system of claim 11, further comprising instructions configured to:

include the first TLC normalized signal in a signal sequence;

determine that the second normalized signal meets a threshold level of temporal similarity to the first normalized signal;

determine that the second normalized signal meets a threshold level of spatial similarity to the first normalized signal; and

include the second normalized signal in the signal sequence.

14. The system of claim 13, wherein instructions configured to aggregate the first single source probability with the second single source probability comprise instructions configured to derive features of the signal sequence.

15. The system of claim 14, further comprising instructions configured to derive features of the signal sequence, including deriving one or more of:

a percentage, a count, a histogram, or a duration.

16. The system of claim 13, wherein instructions configured to determine that the second normalized signal satisfies the threshold level of spatial similarity to the first normalized signal comprise instructions configured to determine that first location dimension and the second location dimension are within a specified distance of one another.

17. The system of claim 13, wherein instructions configured to determine that the second normalized signal satisfies the threshold level of spatial similarity to the first normalized signal comprise instructions configured to determine that first location dimension and the second location dimension are within a specified distance of one another.

18. The system of claim 13, further comprising instructions configured to determine that the second normalized signal is not a duplicate of the first normalized signal prior to including the second normalized signal in the signal sequence.

19. The system of claim 11, wherein instructions configured to receive a first normalized signal comprise instructions configured to receive the first normalized signal corresponding to one of:

a social post with geographic content, a social post without geographic content, an image from a camera feed, a 911 call, weather data, IoT device data, satellite data, satellite imagery, a sound clip from a listening device, data from air quality sensors, a sound clip from radio communication, crowd sourced traffic information, or crowd sourced road information.

20. The system of claim 11, wherein instructions configured to receive the second normalized signal comprise instructions configured to receive the second normalized signal corresponding to a different one of:

a social post with geographic content, a social post without geographic content, an image from a traffic camera feed, a 911 call, weather data, IoT device data, satellite data, satellite imagery, a sound clip from a listening device, data from air quality sensors, a sound clip from radio communication, crowd sourced traffic information, or crowd sourced road information.