Automotive occupancy sensor gray zone neural net conditioning process for airbag deployment systems
First Claim
1. A method of neural net conditioning in an airbag occupancy sensor system for a vehicle interior, said system having a plurality of sensors and at least one sensor signal processing algorithm including a neural net, comprising in any operative sequence the steps of:
- a) establishing a keep out zone;
b) establishing an occupant zone;
c) defining between said keep out and occupancy zones an intermediate gray zone;
d) determining which signals from said sensors are from objects in said gray zone;
e) selecting at least a portion of said signals from objects in said gray zone; and
f) discarding said selected signals when training said neural net to recognize during normal operation at least one of occupant nature, location, or combinations thereof.
1 Assignment
0 Petitions
Accused Products
Abstract
An automotive occupancy sensing system and method for use in conjunction with airbag deployment systems, by which occupant nature, location and motion parameters within the vehicle interior are determined by ultrasound (US) and/or infrared sensors. (IS). Criteria for airbag disablement or airbag modified/partial deployment are used to determine whether appropriate disablement or modified/partial deployment control signals are transmitted to the vehicle airbag deployment system. More particularly the system establishes a Keep Out Zone (KOZ) within the vehicle interior relative to the dashboard or instrument panel and determines actual or imminent incursions by occupants into the KOZ to produce a Keep Out Zone Incursion (KOZI) signal. The system is also characterized by establishing a “Gray Area” bounding the KOZ. Sensor signals from objects/persons in the Gray Zone area selectively used or discarded as part of the training process for a neural net AOS process, and thereby detection performance is substantially enhanced.
7 Citations
11 Claims
-
1. A method of neural net conditioning in an airbag occupancy sensor system for a vehicle interior, said system having a plurality of sensors and at least one sensor signal processing algorithm including a neural net, comprising in any operative sequence the steps of:
-
a) establishing a keep out zone;
b) establishing an occupant zone;
c) defining between said keep out and occupancy zones an intermediate gray zone;
d) determining which signals from said sensors are from objects in said gray zone;
e) selecting at least a portion of said signals from objects in said gray zone; and
f) discarding said selected signals when training said neural net to recognize during normal operation at least one of occupant nature, location, or combinations thereof. - View Dependent Claims (2, 3, 4, 5)
-
-
6. A method of neural net conditioning in an airbag occupancy sensor system for a vehicle interior, said system having a plurality of sensors and at least one sensor signal processing algorithm including a neural net, comprising in any operative sequence the steps of:
-
a) establishing a keep out zone;
b) establishing a non-keep out zone;
c) defining between said keep out and non-keep out zones an intermediate gray zone;
d) selecting a representative known scenario training set selected from known seat empty and seat occupancy scenarios which include scenarios involving each of said zones;
f) receiving and processing signals from said zones, which processing includes discard of information from the gray zone signals that is ambiguous with respect to classification of occupancy state; and
g) extracting from the training set sensor output data on such features as carry actual classification information. - View Dependent Claims (7, 8, 9, 10, 11)
a) training the neural network using known scenarios;
b) testing the neural network performance against a independent test set comprising a different set of known scenarios; and
c) repeating the conditioning and training steps until a selected classification accuracy has been achieved.
-
-
8. The method of claim 7, wherein said at least one sensor signal processing algorithm includes at least one sensor fusion algorithm.
-
9. Method as in claim 8 wherein a fusion algorithm executes the neural network.
-
10. The method of claim 6, wherein said at least one sensor signal processing algorithm includes at least one sensor fusion algorithm.
-
11. Method as in claim 10 wherein a fusion algorithm executes the neural network.
Specification