Method and apparatus for sensing a vehicle crash

Claim 1

In a motor vehicle having a deployable occupant protection apparatus for protecting at least one occupant of said vehicle in a crash, a sensor system for initiating deployment of said occupant protection apparatus comprising: (a) sensor means for sensing accelerations of said vehicle and producing an analog signal representative thereof, (b) mounting means for mounting said sensor means onto said vehicle; (c) converting means for receiving said analog signal and for converting said analog signal into a digital signal; (d) processing means receiving said digital signal and comprising pattern recognition means for determining if said digital signal contains a pattern characteristic of a vehicle crash requiring occupant protection, said processing means producing a deployment signal upon a determination by said pattern recognition means that said digital signal contains the pattern characteristic of a vehicle crash requiring occupant protection, said pattern recognition means comprising a single trained neural network such that the deployment signal is produced solely by the utilization of the single trained neural network; and (e) deployment initiation means coupled to said processing means and responsive to said deployment signal for initiating deployment of said occupant protection apparatus.

Claim 21

A method of obtaining a single neural network for use with a microprocessor-based crash sensor to determine if a deployable occupant protection device in a vehicle should be deployed, comprising the steps of: (a) obtaining digital crash data representative of the vehicle for which the crash sensor is intended to be used, said crash data being obtained from deployment desired crashes, crashes in which deployment is not desired and other events, the combination of all such crashes and events constituting a crash library for said vehicle; (b) designing a candidate pattern recognition algorithm; (c) training said pattern recognition algorithm to categorize the events of said crash library into deployment-desired and deployment-not-desired categories using a pattern recognition computer program and said crash library until categorization errors are reduced to a minimum, resulting in a single trained neural network; (d) testing said trained pattern recognition algorithm using additional crashes and events representative of said vehicle; (e) redesigning said trained pattern recognition algorithm if testing performance is unsatisfactory, and repeating training and testing steps(c) and (d); and (f) outputting from the pattern recognition program the resulting single trained neural network.

Claim 22

A method of obtaining a single neural network for use with a microprocessor-based crash sensor to determine if a deployable occupant protection device in a vehicle should be deployed, comprising the steps of: (a) obtaining digital crash data representative of the vehicle for which the crash sensor is intended to be used, said crash data being obtained from deployment desired crashes, crashes in which deployment is not desired and other events, the combination of all such crashes and events constituting a crash library for said vehicle; (b) designing a candidate pattern recognition algorithm; (c) training said pattern recognition algorithm to categorize the events of said crash library into deployment-desired and deployment-not-desired categories using a pattern recognition computer program and said crash library until categorization errors are reduced to a minimum, resulting in a single trained neural network; (d) testing said trained pattern recognition algorithm using additional crashes and events representative of said vehicle; (e) redesigning said trained pattern recognition algorithm if testing performance is unsatisfactory, and repeating training and testing steps(c) and (d); and (f) outputting from the pattern recognition program the resulting single trained neural network.

Claim 23

A method for sensing a crash of a vehicle to determine if a deployable occupant protection device in the vehicle should be deployed, comprising the steps of: (a) obtaining an acceleration signal from an accelerometer mounted on the vehicle; (b) converting said acceleration signal into digital time series data; (c) entering said digital time series data into a first series of input nodes of a neural network; (d) performing a mathematical operation on said data from each of said first series of input nodes and inputting said operated-on data into a second series of nodes wherein the operation performed on said data from each of said first series of input nodes prior to inputting the operated-on data to the second series of nodes is different from the operation performed on the data from the others of said first series of input nodes; (e) combining the operated-on data from all of the input nodes in said first series of input nodes into each node in said second series of nodes to form a value at each node in said second series of nodes; (f) performing a mathematical operation on each of said values on said nodes in said second series of nodes and inputting said operated-on data into an output series of nodes wherein the operation performed on the value on each of said nodes in said second series of nodes prior to inputting the operated-on value to said output series of nodes is different from that operation performed on the values on others of said nodes in said second series of nodes; (g) combining the operated-on data from all of the nodes in said second series of nodes into each node of said output series of nodes to form a value at each node of said output series of nodes; and, (h) deploying the occupant protection device if the value on one node of said output series of nodes is within a range signifying that a crash requiring the deployment of the occupant protection device is underway.

Claim 24

A method for sensing a crash of a vehicle to determine if a deployable occupant protection device in the vehicle should be deployed, comprising the steps of: (a) obtaining an acceleration signal from an accelerometer mounted on the vehicle; (b) converting said acceleration signal into digital time series data; (c) entering said digital time series data into a first series of input nodes of a neural network; (d) performing a mathematical operation on said data from each of said first series of input nodes and inputting said operated-on data into a second series of nodes wherein the operation performed on said data from each of said first series of input nodes prior to inputting the operated-on data to the second series of nodes is different from the operation performed on the data from the others of said first series of input nodes; (e) combining the operated-on data from all of the input nodes in said first series of input nodes into each node in said second series of nodes to form a value at each node in said second series of nodes; (f) performing a mathematical operation on each of said values on said nodes in said second series of nodes and inputting said operated-on data into an output series of nodes wherein the operation performed on the value on each of said nodes in said second series of nodes prior to inputting the operated-on value to said output series of nodes is different from that operation performed on the values on others of said nodes in said second series of nodes; (g) combining the operated-on data from all of the nodes in said second series of nodes into each node of said output series of nodes to form a value at each node of said output series of nodes; and, (h) deploying the occupant protection device if the value on one node of said output series of nodes is within a range signifying that a crash requiring the deployment of the occupant protection device is underway.