Method and apparatus for detecting key-on and key-off states using time-to-frequency transforms

• US 9,248,789 B2
• Filed: 01/08/2014
• Issued: 02/02/2016
• Est. Priority Date: 01/08/2014
• Status: Active Grant

• Current Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Original Assignee: Verizon Patent and Licensing Incorporated (Verizon Communications Inc.)
• Inventors: Quintero, Francisco
• Primary Examiner(s): Badii, Behrang
• Assistant Examiner(s): GREENE, DANIEL LAWSON
• Application Number:
  US14/150,018
• Publication Number:
  US 20150191138A1
• Time in Patent Office: 755 Days
• Field of Search: 701/39, 701/33.9
• US Class Current: 1/1
• CPC Class Codes: B60R 16/0231 Circuits relating to the driving or the functioning of the vehicle monitoring tyres B60C23/00; indicating overspeed B60K31/00; for dash boards B60K37/00, B60Q3/10; for indicating emergencies B60Q1/52; brake control systems B60T; registering or indicating the working of vehicles G07C5/00; measuring distance G01C, e.g. combinations of speed and distance G01C23/00; engine indicators G01L; measuring speed or acceleration G01P
• CPC Class Codes: B60W 2520/10 Longitudinal speed

Abstract

A key state detector for a vehicle collects a sequence of battery voltage samples and applies a time-domain to frequency-domain transform (TFT) to the collected samples. The results of the TFT are then applied to an artificial neural network (ANN) to determine if they represent a key-on or key-off state. The ANN is trained based on data collected from the vehicle and is periodically retrained so that the detection of key-on and key-off states conforms to the particular vehicle and tracks the aging of vehicle components.

Claims

1. A method comprising:

collecting, by a processor, a sequence of battery voltage samples for a vehicle,

the collected sequence of battery voltage samples being stored in a memory associated with the processor;

calculating, by the processor, a time-domain to frequency-domain transform (TFT) for the collected sequence of battery voltage samples to produce a set of frequency domain features; and

causing, by the processor, a key state of the vehicle to be determined using the set of frequency domain features,

the key state indicating whether the vehicle is active or whether the vehicle is inactive.

2. The method of claim 1, wherein causing the key state of the vehicle to be determined includes applying the set of frequency domain features to an artificial neural network (ANN) to determine the key state of the vehicle,

the method further comprising:

for a particular number of key-state determinations prior to applying the set of frequency domain features to the ANN to determine the key state;

determining the key state of the vehicle by comparing a battery voltage to a threshold;

collecting a first sequence of battery voltage samples for the vehicle;

calculating the TFT for the collected first sequence of battery voltage samples to produce a first set of frequency domain features; and

storing the first set of frequency domain features with respective determined key state as training data; and

after the particular number of key state determinations, training the ANN using the training data.

3. The method of claim 1, wherein calculating the TFT includes calculating at least one of a Fourier transform, a fast Fourier transform (FFT), a wavelet transform, a discrete wavelet transform (DWT), a discrete cosine transform (DCT) or a Hadamard transform.

4. The method of claim 1, wherein causing the key state of the vehicle to be determined includes applying the set of frequency domain features to an artificial neural network (ANN) to determine the key state of the vehicle,

the method further comprising:

confirming that the key state, determined by the ANN, is correct by monitoring at least one of a movement of the vehicle or a speed of the vehicle; and

when the key state is not confirmed, toggling the key state.

5. The method of claim 4, wherein confirming the key state includes:

confirming the key state by monitoring the movement of the vehicle using an accelerometer or by monitoring the speed of the vehicle using a signal provided by an on-board diagnostic (OBD) system of the vehicle.

6. The method of claim 4, wherein the key state is confirmed by monitoring the movement and the speed of the vehicle using a Global Positioning System receiver.

7. The method of claim 4, further comprising:

storing, as training data, each set of frequency domain features and a respective determined key state; and

training the ANN using the training data after a particular amount of training data has been stored.

8. The method of claim 7, wherein the ANN is a one hidden layer feed-forward network, and

wherein training the data includes:

applying the training data to the ANN, and

adjusting coefficients and biases of the ANN based on the applied training data to reduce differences between key states detected by the ANN and the respective determined key state.

9. A system comprising:

an analog-to-digital converter (ADC) to sample a battery voltage of a vehicle at particular intervals to collect respective sets of digitized battery voltage samples;

a memory coupled to the ADC to store the collected respective sets of digitized battery voltage samples; and

a processor configured to;

calculate a time-domain to frequency-domain transform (TFT) for each set of digitized battery voltage samples, of the collected respective sets of digitized battery voltage samples, to produce respective sets of frequency domain features, and

cause a key state of the vehicle to be determined using the respective sets of frequency domain features,

the key state indicating whether the vehicle is active or whether the vehicle is inactive.

10. The system of claim 9, wherein, when causing the key state of the vehicle to be determined, the processor is to cause an artificial neural network (ANN), coupled to the processor, to sequentially process the respective sets of frequency domain features to determine the key state of the vehicle, and

wherein the processor is further configured to:

prior to causing the ANN to sequentially process the respective sets of frequency domain features to determine the key state;

determine the key state of the vehicle by comparing a battery voltage to a threshold;

collect a first sequence of battery voltage samples for the vehicle;

calculate the TFT for the collected first sequence of battery voltage samples to produce a first set of frequency domain features; and

store the first set of frequency domain features with a respective determined key state as training data in another memory; and

wherein the processor is configured to train the ANN using the training data after a particular amount of training data has been stored.

11. The system of claim 9, further including:

an accelerometer; and

a satellite positioning system (SPS) receiver;

wherein, when causing the key state of the vehicle to be determined, the processor is to cause an artificial neural network (ANN), coupled to the processor, to sequentially process the respective sets of frequency domain features to determine the key state of the vehicle;

wherein the processor is further configured to;

confirm that each key state, determined by the ANN, is correct by monitoring at least one of a movement or a speed of the vehicle using at least one of the accelerometer or the SPS receiver, and

when the key state is not confirmed, toggle the key state.

12. The system of claim 11, wherein, when calculating the TFT, the processor is configured to calculate at least one of a Fourier transform, a fast Fourier transform (FFT), a wavelet transform, a discrete wavelet transform (DWT), a discrete cosine transform (DCT) or a Hadamard transform.

13. The system of claim 11, further comprising:

another memory, coupled to the processor, to store, as training data, each set of frequency domain features and a respective determined key state for each confirmed key state,

wherein the processor is configured to train the ANN using the training data after a particular amount of training data has been stored.

14. The system of claim 13, wherein the ANN is a one hidden layer feed-forward network, and

wherein the processor is configured to:

apply the training data to the ANN, and

adjust coefficients and biases of the ANN responsive to the applied training data.

15. A non-transitory computer readable medium storing instructions, the instructions comprising:

one or more instructions which, when executed by a processor, cause flail the processor to;

collect a plurality of sets of battery voltage samples for a vehicle,

the collected plurality of sets of battery voltage samples being stored in a memory associated with the processor;

calculate, for each set of battery voltage samples of the collected plurality of sets of battery voltage samples, a time-domain to frequency-domain transform (TFT) to produce a respective set of frequency domain features; and

cause a key state of the vehicle to be determined using the respective set of frequency domain features,

the key state indicating whether the vehicle is active or whether the vehicle is inactive.

16. The non-transitory computer readable medium of claim 15, wherein the one or more instructions to cause the key state of the vehicle to be determined include one or more instructions to apply the respective set of frequency domain features to an artificial neural network (ANN) to determine the key state of the vehicle, and

wherein the instructions further comprise:

one or more instructions which, when executed by the processor, cause the processor to;

confirm that the key state, determined by the ANN, is correct by monitoring at least one of a or a speed of the vehicle; and

when the key state is not confirmed, toggle the key state.

17. The non-transitory computer readable medium of claim of claim 15, wherein the one or more instructions to cause the key state of the vehicle to be determined include one or more instructions to apply the respective set of frequency domain features to an artificial neural network (ANN) to determine the key state of the vehicle,

wherein the instructions further comprise:

one or more instructions which, when executed by the processor, cause the processor to;

determine the key state of the vehicle by comparing a battery voltage to a threshold, prior to applying the respective set of frequency domain features to the ANN;

collect a first sequence of battery voltage samples for the vehicle prior to applying the respective set of frequency domain features to the ANN;

calculate the TFT for the collected first sequence of battery voltage samples to produce a first set of frequency domain features prior to applying the respective set of frequency domain features to the ANN;

store the first set of frequency domain features with respective determined key state as training data prior to applying the respective set of frequency domain features to the ANN; and

train the ANN using the training data.

18. The non-transitory computer readable medium of claim 15, wherein the one or more instructions to calculate the TFT include one or more instructions which, when executed by the processor, cause the processor to calculate at least one of a Fourier transform, a fast Fourier transform (FFT), a wavelet transform, a discrete wavelet transform (DWT), a discrete cosine transform (DCT) or a Hadamard transform.

19. The non-transitory computer readable medium of claim 15, wherein the one or more instructions to cause the key state of the vehicle to be determined include one or more instructions to apply the respective set of frequency domain features to an artificial neural network (ANN) to determine the key state of the vehicle, and

wherein the instructions further comprise:

one or more instructions which, when executed by the processor, cause the processor to;

store, as training data, each set of frequency domain features and a respective determined key state; and

train the ANN using the training data after a particular amount of training data has been stored.

20. The non-transitory computer readable medium of claim 19, wherein the instructions further comprise one or more instructions which, when executed by the processor, cause the processor to implement the ANN as a one hidden layer feed-forward network.