Claims:We claim:
1. A device (100) to determine type of a vehicle (104), said device (100) comprises a controller (110) adapted to:
receive input signals from an accelerometer (108);
compute a group of characteristic features from said measured input signals,
process said group of characteristic feature using at least one identifier module (112), and
determine at least one of a curb weight and a type of said vehicle (104) based on output of said identifier module (112).

2. The device (100) as claimed in claim 1, wherein a first group of characteristic features are selected from an average acceleration in the three axes, an average of magnitude of the input signal value in the three axes, a standard deviation of the input signal value in the three axes, a Fast Fourier Transform (FFT) of the input signal values in the three axes, a magnitude of the input signals in a horizontal plane of said accelerometer (108), a maximum of the input signal value in each of the three axes and a rate of change of input signal values in each of the three axes.

3. The device (100) as claimed in claim 1, wherein a second group of characteristic features are selected from a standard deviation of the magnitude of input signal value in the three axes, a mean of the magnitude of input signal values in the three axes, a FFT peak comprising an index of the highest FFT value, a FFT ratio comprising a ratio between the largest and a second largest FFT values, a FFT value distribution, an average acceleration in the three axes, an average of magnitude of the input signal values, a maximum acceleration value and an minimum acceleration value in the three axes.

4. The device (100) as claimed in claim 1, wherein said device (100) is selected from group comprising a smartphone, a wearable device, a laptop, a tablet and a cloud server (116).

5. The device (100) as claimed in claim 1, wherein said accelerometer (108) is in any one of said communication unit (102) and a sensor unit (106).

6. A method of determining type of vehicle (104), said method comprising the steps of:
measuring input signals from an accelerometer (108);
computing a group of characteristic features from the measured input signals;
processing said group of characteristic features using at least one identifier module (112), and
determining at least one of a curb weight and a type of said vehicle (104) based on output of said identifier module (112).

7. The method as claimed in claim 6, wherein a first group of characteristic features are selected from an average acceleration in the three axes, an average of magnitude of the input signal value in the three axes, a standard deviation of the input signal value in the three axes, a Fast Fourier Transform (FFT) of the input signal values in the three axes, a magnitude of the input signals in a horizontal plane of said accelerometer (108), a maximum of the input signal value in each of the three axes and a rate of change of input signal values in each of the three axes.

8. The method as claimed in claim 6, wherein a second group of characteristic features are selected from a standard deviation of the magnitude of input signal value in the three axes, a mean of the magnitude of input signal values in the three axes, a FFT peak comprising an index of the highest FFT value, a FFT ratio comprising a ratio between the largest and a second largest FFT values, a FFT value distribution, an average acceleration in the three axes, an average of magnitude of the input signal values, a maximum acceleration value and an minimum acceleration value in the three axes.

9. The method as claimed in claim 6, wherein type of said vehicle (104) is determined by a device (100), said device (100) is selected from group comprising a smartphone, a wearable device, a laptop, a tablet and a cloud server (116).

10. The method as claimed in claim 6, wherein said accelerometer (108) is in any one of said device (100) and a sensor unit (106).
, Description:Complete Specification:
The following specification describes and ascertains the nature of this invention and the manner in which it is to be performed:

Field of the invention:
[0001] The present invention relates to a system and method for determining type of vehicle.

Background of the invention:
[0002] According to a patent literature US 2013/0035829, a use of on-vehicle accelerometer to estimate vehicle grade and mass while vehicle is in motion is disclosed. A system includes a grade estimation module that receives an accelerometer value and generates a grade estimate based on the accelerometer value, wherein the accelerometer value corresponds to acceleration of a vehicle and the grade estimate corresponds to a grade of the vehicle. A mass estimation module receives the accelerometer value and generates a mass estimate based on the accelerometer value, wherein the mass estimate corresponds to a mass of the vehicle. A shift control module at least one of selects and adjusts one of a plurality of shift schedules based on at least one of the grade estimate and the mass estimate and controls a transmission of the vehicle based on the one of the plurality of shift schedules.

Brief description of the accompanying drawings:
[0003] An embodiment of the disclosure is described with reference to the following accompanying drawing,
[0004] Fig. 1 illustrates a device to determine a type of vehicle, according to an embodiment of the present invention, and
[0005] Fig. 2 illustrates a method for determining a type of the vehicle, according to the present invention.

Detailed description of the embodiments:
[0006] Fig. 1 illustrates a device to determine a type of vehicle, according to an embodiment of the present invention. The device 100 comprises a controller 110 adapted to receive input signals from an accelerometer 108, and compute a group of characteristic features from the measured/received input signals. The controller 110 then processes the group of characteristic feature using at least one identifier module 112. The controller 110 determines at least one of a curb weight and a type of the vehicle 104 based on the output of the identifier module 112. The characteristic feature or feature vectors are those that most impact the classification of the curb weight of the vehicle 104. The type of the vehicle 104 corresponds to the make, model and brand of the vehicle 104.

[0007] The accelerometer 108 used is a multi-axis accelerometer 108 such as tri-axial, 6-axis, 9 axis etc. In another embodiment, the along with the input signals from the accelerometer 108, an input from gyroscope (not shown) is also taken. In addition, the raw sensing data of the accelerometer 108 / gyro are also logged at a sampling rate, for example, of 1000 samples per second for the first initial minutes of at the start of every trip of the vehicle 104. Within this time, the curb weight of the vehicle 104 is determined, and from the curb weight, a model / make of the vehicle 104 is determined. In yet another embodiment, the accelerometer 108 is part of an Inertial Measurement Unit (IMU). The device 100 is also enabled to use the input signals from IMU which comprises the gyroscope and the accelerometer, or independently from the gyroscope and the accelerometer 108.

[0008] The feature vectors/parameters that most impact the curb weight of the vehicle 104 is derived or computed by the controller 110, accordingly the identifier module 112 is formed and stored in the memory element of the controller 110. Based on the experiments, the primary feature that most impacted by the curb weight is found to be the vibration experienced by the vehicle 104. The vibrations manifest at changes in the three axes accelerometer values when sampled at a high frequency rate. To capture the data related to vibration, feature vectors are used. Broadly, the features vectors are classified as time domain features and frequency domain features. The features evaluated in the time-domain include mean, magnitude, standard deviation, mean crossing rate, and covariance. The features evaluated in the frequency-domain include entropy, kurtosis, skewness, the highest magnitude frequency, the magnitude of the highest magnitude frequency, and the ratio between the largest and the second largest Fast Fourier Transform (FFT) values. The features in the frequency domain are calculated over frequency domain coefficients on each window of 512, 1024 and 2048 samples. The number of samples are provided for example and not limited thereto.

[0009] In simple words, the principle used for classification is that the vibrations experienced by the vehicle 104 with different loading (weight of the vehicle 104 which includes the weight of the occupants/driver) is different, i.e. when the vehicle 104 starts from stationary position or while experiencing braking. The present invention uses the vibration parameters to classify/determine the curb weight of the vehicle 104, and then to identify the vehicle 104 itself.

[0010] According to an embodiment of the present invention, a first group of characteristic features comprises an average acceleration in the three axes, an average of magnitude of the input signal value in the three axes, a standard deviation of the input signal value in the three axes, a Fast Fourier Transform (FFT) of the input signal values in the three axes, a magnitude of the input signals in a horizontal plane of said accelerometer 108, a maximum of the input signal value in each of the three axes and a rate of change of input signal values in each of the three axes.

[0011] According to another embodiment of the present invention, a second group of characteristic features are selected from a standard deviation of the magnitude of input signal value in the three axes, a mean of the magnitude of input signal values in the three axes, a FFT peak comprising an index of the highest FFT value, a FFT ratio comprising a ratio between the largest and a second largest FFT values, a FFT value distribution, an average acceleration in the three axes, an average of magnitude of the input signal values, a maximum acceleration value and an minimum acceleration value in the three axes. The FFT peak indicates the dominated frequency of the corresponding mode.

[0012] The controller 110 computes the characteristic features either from the first group or the second group or combination thereof, from the real-time measured signals from the accelerometer 108.

[0013] In an embodiment, the identifier module 112 is prepared/formed and stored in a memory element of the controller 110. The identifier module 112 is formed and trained using Artificial Intelligence (AI) and Machine Learning (ML) algorithms. The identifier module 112 is formed based on numerous data sets collected from the accelerometer 108 installed in various type of vehicles 104. For each vehicle 104, a true value is computed and stored in the identifier module 112. The memory element of the controller 110 stores at least one identifier module 112 formed based on the different AI and ML algorithms.

[0014] The identifier modules 112 are created according to model complexity and requirement. A model of low complexity model with lesser number of feature vectors may under fit the data, whereas a high complexity model may suffer from overfitting. A few classifier algorithms are considered for forming identifier module 112, for example Decision Tree, AdaBoost, and Support Vector Machines (SVM). The classifier algorithm are not restricted thereto. All three algorithms independently classify the type of vehicle 104. However, the final classification is based on (but not limited to) a majority vote of the predicted class.

[0015] In one embodiment of the present invention, a sensor unit 106 comprising the accelerometer 108 is installed in the vehicle 104. The sensor unit 106 is dedicated to carry the accelerometer 108. The sensor unit 106 is connectable to the device 100 through either wired or wireless means as known in the art, such as Universal Serial Bus (USB), Type-C, Bluetooth™, Infrared, Wi-Fi, cellular/telecommunication networks such as 2G, 3G, 4G, 5G, etc. The sensor unit 106 is either self-powered having an internal replaceable/chargeable battery or is plugged into a power socket of the vehicle 104 such as a cigar lighter. In another embodiment, the device 100 (specifically a communication unit 102) itself comprises the accelerometer 108. Thus, the accelerometer 108 is in any one of the communication unit 102 and the sensor unit 106. The accelerometer 108 is shown to be in the communication unit 102 for simplicity and must not be limited thereto as described above.

[0016] According to the present invention, a method of working of the device 100 is described. First, a communication is established between the accelerometer 108 and the device 100. The accelerometer 108 with which the communication is established is in the communication unit 102 or is in the sensor unit 106. Next, when the vehicle 104 is driven by the driver, all the measured input signals are recorded by the communication unit 102. The controller 110 on receiving the required input signals, computes the characteristic features based on either of the first group and the second group. The controller 110 uses the identifier module 112 and computes a confidence score for a type of vehicle 104. Based on the confidence score, the type of the vehicle 104 is identified.

[0017] An example of determining/identifying the type of the vehicle 104 is disclosed. A table 114 shows multiple identifier module 112 IM1, IM2 and IM3. The three identifier modules 112 are stored in the memory element of the controller 110 which are created by various AI and ML algorithms as disclosed above. While only three identifier modules 112 are shown, but it is possible to use at least one and more than three as well. Consider the controller 110 determines the type of the vehicle 104 using IM1. The confidence score is 0.63 (out of 1) which indicates the confidence of the identifier module 112 for the detected type of vehicle 104 V1. The confidence score is computed by the controller 110 based on comparison of the output value of the IM1 with the true value (empirically or experimentally calculated) stored for the V1. Similarly, the confidence score of the other possible type of vehicles 104 is determined as 0.3 for V2, 0.95 for V3 and 0.1 for V4. The values of other cells are dashed to indicate random numbers. The controller 110 uses either the first group or the second group of characteristic features to calculate the confidence score. Assuming only one identifier module 112 is used, then based on the score, the type of the vehicle 104 is detected to be of the highest score, i.e. V3. Alternatively, the confidence score which is above a threshold is selected for the determination of the type of vehicle 104.

[0018] In another case, the controller 110 comprises three identifier modules 112 and the confidence score of the all the three identifier modules 112 are used to determine the type of the vehicle 104. Now, in yet another case it is possible that the confidence score of two identifier modules 112, say of IM1 and IM2 for two different vehicles 104, say V2 and V3 are identical. Here, there exist a conflict or confusion. In such a scenario, the controller 110 again evaluates through the respective identifier modules 112 IM1 and IM2, but with removal of the impactful/influential features from the used group of characteristic features. This leads to a different score, and the controller 110 finalizes the type of the vehicle 104 based on the highest score. Though, the confusion or conflict is explained with respect to the different identifier modules 112, but the confusion or conflict is possible to occur within the confidence score of same identifier module 112 as well, which is also resolved by the same method.

[0019] According to an embodiment of the present invention, the device 100 is selected from group comprising a communication unit 102 and a cloud server 116. The communication unit 102 is selected from a smartphone, a wearable device, a laptop and a tablet.

[0020] In an embodiment, the device 100 is the cloud server 116. Now, w.r.t the Fig. 1, a control unit (not shown) of the cloud server 116 performs the processing. The communication unit 102 is just used to transfer the measured input signals from the accelerometer 108 to the cloud server 116. The identifier modules 112 are stored in the cloud server 116. The communication unit 102 establishes the connection with the cloud server 116 and transmits the real time input signal measured from the accelerometer 108. The computation is performed in the cloud server 116 to derive the characteristic feature either as per the first group or the second group. The cloud server 116 then processes the computed or calculated characteristics features through the identifier module 112. Finally, the cloud server 116 determines the at least one of the curb weight and the type of the vehicle 104 based on the output of the identifier module 112.

[0021] In still another embodiment, the determination of the type of the vehicle 104 is performed partially by the controller 110 of the communication unit 102 and the cloud server 116. In an alternative, a default mode is set to cloud server 116 to determine the type of the vehicle 104. However, in the absence of connectivity to the cloud server 116, due to network error scenarios, the controller 110 itself performs the determination of the type of the vehicle 104 and updates the cloud server 116 once the network is connected/available.

[0022] The type of the vehicle 104 is determined by the controller 110 and displayed in the device 100. In an embodiment, the step of determining the type of the vehicle 104 is triggered through a mobile application installed in the communication unit 102. The mobile application is opened and a virtual button is pressed to initiate the determination of the type of the vehicle 104.

[0023] Fig. 2 illustrates a method for determining a type of the vehicle, according to the present invention. The method comprises plurality of steps, of which a step 202 comprises, measuring input signals from the accelerometer 108. A step 204 comprises computing a group of characteristic features from the measured input signals. A step 206 comprises processing the characteristic features using at least one identifier module 112. A step 208 comprises determining at least one of the curb weight and the type of the vehicle 104 based on output of the identifier module 112.

[0024] The steps 202 through 208 are executed by any one of the controller 110 of the communication unit 102 or the cloud server 116. In the embodiment of the cloud server 116, the communication unit 102 facilitates the transmission of the input signals measured from the accelerometer 108.

[0025] The first group of characteristic features are selected from the average acceleration in the three axes, the average of magnitude of the input signal value in the three axes, the standard deviation of the input signal value in the three axes, the Fast Fourier Transform (FFT) of the input signal values in the three axes, the magnitude of the input signals in a horizontal plane of the accelerometer 108, the maximum of the input signal value in each of the three axes and a rate of change of input signal values in each of the three axes.

[0026] The second group of characteristic features are selected from, the standard deviation of the magnitude of input signal value in the three axes, the mean of the magnitude of input signal values in the three axes, the FFT peak comprising an index of the highest FFT value, the FFT ratio comprising the ratio between the largest and the second largest FFT values, the FFT value distribution, the average acceleration in the three axes, the average of magnitude of the input signal values, the maximum acceleration value and the minimum acceleration value in the three axes.

[0027] The type of the vehicle 104 is determined by the device 100. The device 100 is selected from group comprising the communication unit 102 and the cloud server 116. The communication unit 102 is any one selected from a smartphone, a wearable device, a laptop and a tablet.

[0028] According to embodiments of the present invention, the device 100 and a method is provided to collect data from an accelerometer 108, and process the aggregated data to determine the type of the vehicle 104. In simple words, a device 100 and method for determining type of vehicle 104 from the accelerometer 108 data is disclosed. The identifier modules 112 are made using classification/classifier algorithms offline, and then deployed on the device 100. The present invention identifies the vehicle 104 without any communication with the vehicle 104. The device 100 also makes use of occupant weight (approximate or accurate), which is mapped with the mobile/user application installed in the device 100. The determined type of the vehicle 104 is used in the insurance sector among other uses. As an example, the type of the vehicle 104 is used to determine the premium to the paid for insuring the asset. Alternatively, the type of vehicle 104 is used to validate the vehicle information submitted by an owner. Apart from the accelerometer 108, additional data such as weather information, terrain conditions along the route is also captured. Thus, the present invention is made possible by the use of feature vectors that most impact the classification of the curb weight of the vehicle 104. Further, the acceleration parameters and vehicle vibration calibration parameters are used for the formation of the identifier modules 112. An information on road gradient is also used either directly or from the gyroscope and the acceleration data. The present invention determines the type of the vehicle 104 by classifying the aggregated vibration data, road gradient and maps to the curb weight of the vehicle 104.

[0029] It should be understood that embodiments explained in the description above are only illustrative and do not limit the scope of this invention. Many such embodiments and other modifications and changes in the embodiment explained in the description are envisaged. The scope of the invention is only limited by the scope of the claims.