Methods for characterizing a low-impact vehicle collision using high-rate acceleration data

1. A method, comprising:

capturing, by a telematics monitor coupled to a vehicle, high-rate acceleration data, speed data, and global positioning system (GPS) data;

identifying at least one high-rate acceleration trigger event from the high-rate acceleration data;

computing a plurality of feature values for a plurality of features corresponding to the at least one high-rate acceleration trigger event;

when at least some of the plurality of feature values indicate that the at least one high-rate acceleration trigger event comprises a noise event or a harsh braking event, filtering out the at least one high-rate acceleration trigger event from the high-rate acceleration data;

when the at least some of the plurality of feature values indicate that the at least one high-rate acceleration trigger event does not comprise a noise event and does not comprise a harsh braking event;

computing a plurality of feature scores corresponding to the plurality of feature values;

computing an acceleration context based on the plurality of feature scores;

computing, by a collision detection facility local to the telematics monitor, each of:

the acceleration context based on the high-rate acceleration data;

a pullover context based on the speed data;

a speed context based on the speed data; and

a road context based on the GPS data;

computing, by the collision detection facility, an accident score based on the acceleration context, the pullover context, the speed context, and the road context; and

in response to determining that the accident score indicates a potential collision, triggering at least one action for responding to the potential collision.

2. The method of claim 1 , wherein capturing the high-rate acceleration data comprises capturing acceleration data at a rate between 80 Hz and 120 Hz.

3. The method of claim 1 , wherein identifying the at least one high-rate acceleration trigger event from the high-rate acceleration data comprises identifying a plurality of successive acceleration points having a magnitude which is above a trigger threshold.

4. The method of claim 1 , wherein identifying the at least one high-rate acceleration trigger event from the high-rate acceleration data comprises:

identifying a first plurality of successive acceleration points having magnitudes above a trigger threshold;

identifying a second plurality of successive acceleration points having magnitudes above the trigger threshold; and

when a first acceleration point of the second plurality of successive acceleration points has a first acceleration point timestamp that is within a predetermined time window after a last acceleration point timestamp of a last acceleration point in the first plurality of successive acceleration points, combining the first plurality of successive acceleration points and the second plurality of successive acceleration points to form the at least one high-rate acceleration trigger event.

5. The method of claim 1 , wherein computing the plurality of feature values for the plurality of features corresponding to the at least one high-rate acceleration trigger event comprises:

defining an impact window for the at least one high-rate acceleration trigger event;

computing an area-under-the-curve (AUC) for the impact window;

computing a deviation change for the impact window;

determining an impact window length for the impact window;

determining a maximum acceleration point value of the impact window; and

determining a number of zero crossings for the impact window.

6. The method of claim 5 , wherein at least some of the plurality of feature values indicate that the at least one high-rate acceleration trigger event comprises a noise event when:

the AUC comprises a combined AUC for a combined X-direction and Y-direction acceleration profile inside the impact window;

the combined AUC has a lower magnitude than a noise AUC threshold;

the deviation change is greater than a noise deviation change threshold; and

the number of zero crossings is greater than a noise zero crossings threshold.

7. The method of claim 5 , wherein at least some of the plurality of feature values indicate that the at least one high-rate acceleration trigger event comprises a harsh braking event when:

the AUC is more negative than a harsh braking AUC threshold;

the deviation change is less than a harsh braking deviation change threshold; and

the maximum acceleration point value is lower than a low maximum acceleration threshold.

8. The method of claim 5 , wherein computing the plurality of feature scores, comprises:

computing an AUC score from the AUC;

computing a deviation score from the deviation change; and

computing an impact length score from the impact window length.

9. The method of claim 8 , wherein computing the plurality of feature scores further comprises computing a maximum acceleration score.

10. The method of claim 8 , wherein computing the AUC score, comprises:

comparing an X-direction maximum acceleration for an X-direction acceleration profile of the at least one high-rate acceleration trigger event and a Y-direction maximum acceleration for a Y-direction acceleration profile of the at least one high-rate acceleration trigger event;

when the X-direction maximum acceleration is greater than the Y-direction AUC maximum acceleration, computing an X-direction AUC and selecting the X-direction AUC as a selected AUC that is to be mapped to the AUC score; and

when the Y-direction maximum acceleration is greater than the X-direction maximum acceleration, computing a Y-direction maximum acceleration and selecting the Y-direction maximum acceleration to be the selected AUC that is to be mapped to the AUC score.

11. The method of claim 10 , wherein mapping the selected AUC to the AUC score, comprises:

when the selected AUC is lower than a maximum negative AUC or is higher than a maximum positive AUC, setting the AUC score to 1;

when the selected AUC is higher than the maximum negative AUC and is lower than a medium negative AUC, setting the AUC score to:

1−(the selected AUC−the maximum negative AUC)/(the medium negative AUC−the maximum negative AUC);

when the selected AUC is higher than the medium negative AUC and lower than a medium positive AUC, setting the AUC score to 0; and

when the selected AUC is higher than the medium positive AUC and is lower than the maximum positive AUC, setting the AUC score to:

(the maximum positive AUC−the selected AUC)/(the maximum positive AUC−the medium positive AUC).

12. The method of claim 8 , wherein computing the deviation score, comprises:

for each of an X-direction acceleration profile and a Y-direction acceleration profile of the at least one high-rate acceleration trigger event:

computing a first standard deviation for a first plurality of acceleration points within the impact window;

computing a second standard deviation for a second plurality of acceleration points outside the impact window but within a predetermined duration thereof; and

computing the deviation change by dividing the first standard deviation by the second standard deviation;

wherein the deviation change computed for the X-direction acceleration profile is an X-direction deviation change, and the deviation change computed for the Y-direction acceleration profile is a Y-direction deviation change.

13. The method of claim 12 , wherein computing the deviation score further comprises:

selecting a higher deviation change of the X-direction deviation change and the Y-direction deviation change as a selected deviation change;

when the selected deviation change is lower than a minimum deviation change, setting the deviation score to 0;

when the selected deviation change is higher than a maximum deviation change, setting the deviation score to 1; and

when the selected deviation change is higher than the minimum deviation change and lower than the maximum deviation change, setting the deviation score to:

(the selected deviation change−the minimum deviation change)/(the maximum deviation change−the minimum deviation change).

14. The method of claim 8 , wherein determining the impact length score comprises:

when the impact window length is lower than a minimum impact window length or higher than a maximum impact window length, setting the impact length score to 0;

when the impact window length is higher than a low impact window length and lower than a high impact window length, setting the impact length score to 1;

when the impact window length is higher than the minimum impact window length and is lower than the low impact window length, setting the impact length score to:

(the impact window length−the minimum impact window length)/(the low impact window length−the minimum impact window length); and

when the impact window length is higher than the high impact window length and lower than the maximum impact window length, setting the impact length score to:

1−(the maximum impact window length−the impact window length)/(the maximum impact window length−the minimum impact window length).

15. The method of claim 9 , wherein computing the maximum acceleration score comprises:

when the maximum acceleration point value is lower than a low maximum acceleration threshold, setting the maximum acceleration score to 0;

when the maximum acceleration point value is higher than a high maximum acceleration threshold, setting the maximum acceleration score to 1; and

when the maximum acceleration point value is higher than the low maximum acceleration threshold and lower than the high maximum acceleration threshold, setting the maximum acceleration score to:

(the maximum acceleration point value−the low maximum acceleration threshold)/(the high maximum acceleration threshold−the low maximum acceleration threshold).

16. The method of claim 1 , wherein computing the accident score comprises computing a weighted average of the acceleration context, the pullover context, the speed context, and the road context.

17. The method of claim 1 , wherein triggering the at least one action comprises placing the telematics monitor in a data-gathering mode for gathering additional data pertaining to a collision.

18. The method of claim 1 , wherein triggering the at least one action comprises making an emergency call from the telematics monitor.

19. The method of claim 5 , wherein defining the impact window uses a threshold-based method in which:

the impact window starts at a first acceleration point having an acceleration magnitude that is higher than an impact window trigger threshold; and

the impact window ends at a second acceleration point having an acceleration magnitude that is lower than the impact window trigger threshold.