FR3091754A1 - Identification of a motor vehicle by its vibratory characteristics - Google Patents

Abstract

Identification of a motor vehicle by its vibratory characteristics The present invention relates to the identification of a motor vehicle as a function of a vibratory signal that it emits and in which provision is made to constitute a learning base (S1) comprising vibration signals, and a pickup of a current vibratory signal (S2) followed by an implementation of artificial intelligence to compare (S3) the current signal with the signals of the learning base. Depending on a result of the comparison, the current vehicle is determined (S5) as or not corresponding to the vehicle (VEHi) associated with the learning base. In particular, the vibration signals of the learning base have been acquired by the same vibratory signal pickup and are stored (S6) in the learning base in correspondence of a pickup instant. Thus, each time a new vibratory signal is picked up on the vehicle, if this vehicle is identified as being the vehicle associated with the aforementioned learning base, the latter is updated (S7) by replacing the oldest vibration of the learning base by the new vibratory signal picked up. [Fig. 2]

Description

Identification of a motor vehicle by its vibration characteristics

The present invention relates to the identification of a motor vehicle by recognition of vibration and/or sound characteristics of at least part of the vehicle (its engine, an element of its bodywork, or others).

There are known processings of vibration signals (in particular acoustic vibrations) making it possible to recognize the type of engine in a vehicle. More particularly, it can be a question of identifying the make of the engine and its model. With reference to FIG. 1 relating to the prior art, an engine vibration sensor CAP (for example a microphone picking up a sound vibration signal from the engine) transmits a signal VIB of the vibrations picked up to an input interface IN of a processing circuit CT (for example of a computing device such as a laptop computer, a tablet, a smartphone or the like). A processor PROC of the processing circuit CT executes instructions of a computer program stored in a memory MEM of the processing circuit for:
- Compare the VIB signal with typical motor vibration signals, referenced in a database (this database being stored for example in the memory MEM or with a remote server, and for this purpose the processing circuit may comprise in in addition to a communication interface with the aforementioned server),
- Identify in the database the best match between the VIB signal and one of the standard signals, and
- Recognize the type of vehicle engine MOD# from which the VIB signal comes, as corresponding to the engine referenced in the database and associated with the identified type signal. An output interface OUT of the processing circuit can then issue a command to a man/machine interface (not shown) to display, for example, the recognized engine type MOD#.

Such processing thus makes it possible to recognize a given type of engine for a plurality of vehicles equipped with the same type of engine.

Technical problem

However, such processing does not make it possible to recognize a particular vehicle, nor to distinguish this particular vehicle from other vehicles equipped with the same type of engine.

In addition, the vibration of the engine changes as it ages, and the typical signals referenced in the aforementioned database may no longer make it possible to identify a type of engine having undergone this aging.

The present invention improves this situation.

To this end, it proposes a method for identifying a motor vehicle as a function of a vibration signal originating from at least part of the vehicle, the method comprising:
- a constitution of a learning base comprising vibration signals from said part of the vehicle, this learning base thus being associated with said vehicle,
- a capture of a vibration signal from a homologous part of a current vehicle,
- an implementation of an artificial intelligence to compare the vibration signal picked up with the vibration signals of the learning base, and
- Depending on a result of the comparison, a determination of the current vehicle as corresponding or not to the vehicle associated with the learning base.

Specifically :
- the vibration signals of the learning base are acquired by said vibration signal capture and stored in the learning base in correspondence of a capture instant, and
- each time a new vibration signal is picked up on a current vehicle, if the current vehicle is identified as being the vehicle associated with the learning base, the learning base is updated by replacing the vibration signal the oldest of the learning base by the new vibratory signal picked up.

Above, the term "corresponding to a moment of capture" means the fact that a relative seniority of the signals is taken into account to order the signals in the base (such as a memory of the "FIFO" type (for "First In First Out”) which thus forms part of the possible embodiments of the present invention).

Thus, the oldest signals of the base can be obsolete because the vehicle evolves during its use (by aging or other reason), and therefore the vibration of the vehicle also evolves. The obsolete signals of the learning base are in particular replaced by the latest signals which were acquired most recently and which still made it possible to identify the vehicle. The use of the learning base thus updated makes it possible to more robustly recognize the vehicle in its current state.

Thus, the implementation of the invention makes it possible not only to recognize a model of vehicle leaving the factory, but also to recognize this vehicle subsequently (as illustrated by the reference VEHi in the box in dotted lines specific to the invention in Figure 1), despite its aging for example.

In one embodiment, to identify a particular vehicle from a plurality of vehicles:
- a learning base is provided for each vehicle of said plurality of vehicles,
- the capture of a new vibration signal from a part of the current vehicle is followed by a comparison of the new signal captured at each of the learning bases with an estimation of a comparison score for each learning base, and
- the current vehicle is identified as being the vehicle associated with the learning base whose comparison score is the highest, the method continuing with the replacement of the oldest vibration signal of the learning base with the highest score higher by the new signal received.

It is thus possible, by this implementation, to distinguish a plurality of vehicles, for example by uploading various signals picked up to a server storing the aforementioned learning bases (or connected to a storage memory of these bases).

In one embodiment, the part of the vehicle from which the vibration signal originates can be an engine of the vehicle (for example a heat engine), and the aforementioned learning base comprises vibration signals characteristic of engine operating phases, said phases being marked in the signals of the learning base and the artificial intelligence using marks thus associated each with one of said phases.

For example, said phases may comprise at least one acceleration, a neutral point, and a steady state of the engine.

Indeed, one can initially mark (or "tag") different phases in the signals of the learning base, such as for example:
Acceleration and/or deceleration of the motor,
A standstill,
An extinction and/or an ignition of the engine,
Or other.

As a variant, it may be at least one characteristic vibration of another element of the vehicle than the engine, for example a slamming of a door (for example of the driver), or other. Typically, for an electric vehicle, the vibratory signal can characterize for example a braking of the vehicle, or even a suspension of the vehicle, or even a rotation of a wheel of the vehicle, or other.

In one embodiment, the artificial intelligence uses an artificial neural network with deep learning of the learning base.

Here, “deep learning” means in particular the fact that the neural network recognizes in the captured signal not only a type of vibration (for example one particular of the aforementioned phases), but also the vehicle that caused this vibration.

It is thus possible to initially mark, for example, phases of the engine of a vehicle of the same model. The neural network will then be able to identify these phases on a current vehicle to be identified, but will also be able, by deep learning, to recognize the specific vibration of this current vehicle, for each of these phases.

In one embodiment, the vibration signal picked up is acquired by a sensor of a mobile device available to a user of the vehicle, and capable of communicating data from the signal picked up to a remote server storing the learning base (or connected to a memory storing this database).

It may for example be a smartphone comprising such a sensor and which the user carries on him. Nevertheless, in an advantageous variant, it may be a spectacle frame comprising a sensor for implementing the method.

In one embodiment, the mobile device includes an internal battery and a mains power supply connection module for recharging the internal battery, and the mobile device is configured to condition the communication of data from the received signal to the remote server when the internal battery is charging.

Such an embodiment makes it possible to request the mobile device to send data only in an inactive phase of the mobile device (when the user is not particularly using his device) and typically in a home where access to a network communication with the remote server is available.

Above, the vibratory signal picked up can be a sound signal acquired by a microphone-type sensor carried by the mobile device.

In a variant, the vibration signal picked up is acquired by an acceleration sensor intended to be worn by a predetermined user of the vehicle, the method further comprising an implementation of artificial intelligence for:
- to distinguish, in the signals acquired by the sensor, the vibratory signals of the part of the vehicle, from the gesture signals of the user,
- classify the learning base by distinguishing vibratory signals and gestural signals,
- and, in addition to determining the current vehicle as corresponding or not to the vehicle associated with the learning base, determining a correspondence or not between the user from whom the gesture signals in the captured signal originate and the user from whom the hand signals from the learning base.

Thus, such an embodiment where the sensor used is a motion sensor makes it possible both to identify the vehicle but also the user of this vehicle by recognizing gestures characteristic of this user. Such an embodiment makes it possible in particular to implement a service for verifying users authorized to drive the vehicle, for example.

For example, provision can be made for the method to further comprise triggering an alert signal in the event of a mismatch between the user from whom the gesture signals in the captured signal originate and the user from whom the signals originate. gestures of the learning base.

In one embodiment, the aforementioned acceleration sensor can be arranged on a pair of glasses frame intended to be worn by the user from whom the gesture signals of the learning base originate.

Thus, such an arrangement of the sensor can make it possible to recognize characteristic gestures of the user (for example the way in which he raises the glasses on his nose, or even more particularly a specific movement of the head to look at the mirrors, thus also showing that the user is currently driving the vehicle, or a characteristic bodily gesture to get into the vehicle, further signifying that the user is in the vehicle, or the like).

The glasses can be, for example, eyeglasses or even sunglasses with (filtering) lenses, which are intended to be permanently in the vehicle, for example.

Provision can also be made for this sensor to be capable of communicating (via a short-range link of the Bluetooth type or other) with a mobile terminal connected to said learning base (typically via an extended network).

Alternatively again, the learning base can be stored in the terminal itself, the latter then being able to manage the identification of the vehicle without having recourse to a server.

As indicated previously, it is possible, by the implementation of the invention, to identify a vehicle throughout its development, and this also, as presented below, when this vehicle begins to exhibit a failure. More particularly, the present invention makes it possible to identify a specific vehicle by a characteristic vibration, but may additionally make it possible, in a particular embodiment, to identify a failure occurring in the vehicle. In such an embodiment, the method may then further comprise:
- establishment, from the learning base, of a vibration model of said part of the vehicle,
- after capturing the vibration signal from said part of the vehicle, an estimate of a difference between the vibration pattern and the vibration signal captured, and
- If said estimate is greater than a threshold, a triggering of a warning signal of a failure of said part of the vehicle.

Advantageously, the recognition of the vehicle is sufficiently robust, due to the replacement of the old signals by the signals newly acquired in the base, to continue to identify the vehicle despite its evolution and possibly despite a failure. Thus, in the embodiment above, a vibration model is stored in memory (for example a standard model leaving the factory), and if the vibration deviates considerably from this vibration model (possibly for an age of the vehicle below a threshold), then a vehicle failure can be suspected.

In such an embodiment, it is also possible to provide:
- a storage of a plurality of vibration models of homologous parts of vehicles, said homologous parts having respective pre-identified failures, and
- if said estimate of difference between the model and the signal picked up is greater than the threshold, a comparison of the signal picked up with each of the vibration models of homologous parts with respective failures, to identify the failure of said part of the vehicle from which the vibration signal picked up.

It is thus possible to identify in particular the type of failure that the vehicle has. In the example presented here, this characterization of the type of failure is implemented in the event of a significant difference detected between the model and the signal received, so as to avoid performing this test unnecessarily. However, in a variant, this test can be carried out periodically, or at a frequency increasing with the age of the vehicle, or otherwise. Thus, the fact of taking into account that the estimate of difference between the model and the captured signal is greater than the threshold, can be optional.

Furthermore, in the event of identification of a given failure, as the user of the vehicle can also be recognized as described above, it is possible to send him the aforementioned warning signal of a failure of an identified part of the vehicle, for example in the form of a notification on his smartphone, in order to make him aware of the failure in progress, which could cause a breakdown of the vehicle. For this purpose, a database can store the learning base in correspondence of identifiers in particular of:
Smartphone (typically call number),
Vehicle (registration or other),
User name (surname, first name, etc. or company name, company name, etc.)

The present invention also relates to a computer program comprising instructions for the implementation of the method above, when these instructions are executed by a processor of a processing circuit.

By way of example, at least one of the flowcharts presented below with particular reference to FIGS. 2, 3 and 5 can illustrate the general algorithm of such a program.

The present invention also relates to a system for implementing the above method and comprising at least:
- a memory to store the learning base,
- a sensor of said vibratory signals, and
- a processing circuit implementing said artificial intelligence,
the processing circuit being configured to store in the learning base the vibration signals picked up, in correspondence of a moment of picking up, and for, each time a new vibration signal is picked up on a current vehicle, in the event of identification of the current vehicle as being the vehicle associated with the learning base, updating the learning base by replacing the oldest vibration signal of the learning base with the new vibration signal picked up.

An exemplary embodiment of such a system is illustrated in FIG. 4, commented on below.

Moreover, other characteristics and advantages of the invention will appear on reading the detailed description below, and on examining the appended drawings, in which:

Fig. 1

illustrates a principle of identification of a vehicle model as described above;

Fig. 2

illustrates the main steps of a method according to an example embodiment,

Fig. 3

illustrates the main steps of a method according to one embodiment of vehicle recognition among a plurality of vehicles;

Fig. 4

illustrates an exemplary embodiment of a system for implementing the above method;

Fig. 5

illustrates the main steps of a method according to an embodiment where a current user of the vehicle is also recognized,

Fig. 6

illustrates the vibrations of a vehicle engine during different operating phases for recognition of this vehicle from an analysis of these vibrations.

With reference to FIG. 2, a first step S1 consists in holding several learning bases APPRi, each specific to a motor vehicle VEHi, stored in a memory (for example of a server as described later). At step S2, a vibration signal is picked up by a CAP sensor, and this signal is compared by an artificial intelligence to the signals of each learning base APPRi, successively (loop S4 on the index i of the learning bases APPRi, at the “KO” output of the test S3), until identifying in step S5 the vehicle VEHi from which this vibration signal SENS could be picked up. In an exemplary embodiment, an artificial neural network can be implemented to extract from each learning base APPRi a vibration model (or more particularly specific vibration characteristics) specific to each vehicle VEHi, and step S3 can then match a comparison of the newly acquired SENS vibration signal to each of these vibration patterns.

To do this, it is possible to rely, by way of purely illustrative example, on particular phases of operation of the engine of the vehicle (in particular for a heat engine), as illustrated in FIG. 6, representing the evolution of a vibration intensity modulus (over several frequencies) as a function of time over several engine operating phases:
When it is turned on Ca,
At its idle speed Cb,
In DC acceleration,
In steady state in Cf grip,
In Cd deceleration,
At its extinction Ce.

Each of these phases can be "tagged" by an artificial intelligence, with the following possible relevant parameters:
vibration frequencies,
The amplitude associated with each of these frequencies,
And/or others.

For example, deep learning can be implemented (using an artificial neural network) and thus not only can each of these phases be recognized in the vibration signal captured SENS but also the vehicle from which this signal originates can be identified as described below.

Referring again to FIG. 2, once a correspondence has been identified between the signal SENS and one of the learning bases (or its model), the corresponding vehicle VEHi is then determined (step S5 at the output “ OK” of test S3) and, in particular, the newly sensed signal SENS is stored in the learning base APPRi of this vehicle VEHi (step S6). This storage can optionally be operated as a replacement for an oldest signal picked up from the base in step S7 (for example according to a mechanism of the FIFO (or “First In First Out”) type). update of the base to follow the evolution of the vehicle, while also keeping a reasonable memory size for storage.

Reference is now made to FIG. 3 to detail an embodiment relating to the use of several learning bases to identify the one corresponding to the vehicle from which the newly sensed vibration signal originates. In addition to the common steps S4 and S2 of the preceding FIG. 2, to implement step S3 more precisely, a score SCOi is calculated for comparing the newly sensed signal SENS with the content of each learning base APPRi at step S21. The maximum score MAX is determined in steps S22 and S23, over the set of N learning bases (test S24). The vehicle VEHi is determined as being the one associated with the learning base APPRi which has the maximum score at step S25, and this learning base is updated at step S26. Step S27 then consists of resetting the MAX value to zero to reiterate the search loop for the maximum score in the example embodiment illustrated in FIG. 3.

As illustrated in FIG. 4, this method above can be implemented by a system comprising, in the example represented, a pair of glasses connected LUN, comprising a movement sensor CAP (for example an acceleration sensor) fixed for example to the frame of the pair of glasses. This sensor can then be arranged to capture vibration data from a part of the vehicle (engine or other), but also from the movements of a user wearing this pair of glasses. Thus, as we will see later, the CAP sensor can allow both the recognition of the vehicle, but also of the user in this vehicle.

The CAP sensor may include, for example, a communication antenna (Bluetooth, Wi-Fi or other) of the signals picked up for storage in the memory (at least temporarily) of a smartphone SM that the system also includes in an exemplary embodiment. This SM smartphone typically includes a BATT power supply battery that can be recharged via a MODC connection module to a SECT mains supply. It will thus be understood that the smartphone itself can communicate the signal data acquired by the CAP sensor, as soon as it is connected for example to the mains via its MODC module. These data can be transmitted via an extended communication network RES to a server SER which stores the learning bases APPRi and which can thus implement the steps described above for comparing the newly picked up signal with the respective contents of the bases of stored learning.

Referring now to FIG. 5, the step S31 which follows the step S2 of picking up the vibration signal SENS can consist in trying to detect a movement of the spectacle wearer LUN. Where appropriate, in step S32, an attempt is made to recognize the wearer of these glasses among several users whose characteristic gestures are stored in a user base DB US. In the event that none of the users listed in the database DB US could be identified, it is possible to generate an alert at step S34, intended for example for the smartphone of a main user whose access number is entered in the database DB US and thus obtained in step S33. This can typically be a user declared as being the main user of the vehicle within the same family, or the same company.

If the current user is well recognized in the DB US base, it is then possible to send him notifications directly on his smartphone (his contact number also being entered in the DB US base), in particular at step S40 described further away.

Moreover, as also illustrated in FIG. 5, once the vehicle has been identified in step S5, the newly sensed vibration signal can be compared to a vibration model MOD of the vehicle (estimated from the learning base and typically given by an artificial neural network for example) to determine whether the difference between the signal SENS and this model MOD is beyond a chosen threshold THR. If this is the case, such a measurement may characterize the presence of a vehicle failure.

Thus, to further characterize this failure, it is possible to determine in step S36 the model of the vehicle which was identified in step S5. It is thus possible to search, as follows, for the type of failure in a DB BKD database of characteristic vibration models in the event of various failures (engine, suspension, braking, etc.) and this for different vehicle models. Once the vehicle model has been identified (step S36), it is possible to compare the newly acquired signal with each of the vibration models in the event of failure of this vehicle model, for example by assigning scores SCOj to each of these comparisons to the step S37. The vibration model in the event of a failure which obtains the maximum score is then identified in step S38. The current failure BKDj can thus be recognized and it is then possible to generate an alert at step S39 intended for the smartphone of the current user of the vehicle, at step S40.

Such an achievement makes it possible to make the current user aware of a particular noise from his vehicle in operation, which he can hear directly, and which, according to the processing presented above, characterizes a particular breakdown which is likely to occur (or will occur). after a determined number of kilometers covered). The alert message on the smartphone of the current user (or alternatively on the smartphone of the aforementioned main user) can thus recommend the replacement of a particular part of the vehicle (tyres, shock absorbers, engine suspension, brake pads or others).

Of course, the present invention is not limited to the embodiments described above by way of example; but it extends to other variants.

Thus, for example, the fact of providing a frame of pairs of glasses to carry the CAP sensor can be optional. Alternatively, the sensor can be a motion sensor of a smartphone or of a connected bracelet or other, which the user wears and is thus recognized. The smartphone can then additionally use a microphone to pick up vibration signals in the form of acoustic signals. Alternatively again, the CAP sensor is installed in the motor vehicle and transmits the signals picked up to the smartphone when the latter is detected in the vehicle. In this embodiment, as the smartphone transmits the data of the signals picked up to the server SER generally with a smartphone identifier, it is possible to notify this smartphone of any alert signal of the aforementioned type. Of course, the term “smartphone” here means any connected device such as a tablet, laptop, connected bracelet or other.

Claims (14)

Method for identifying a motor vehicle as a function of a vibratory signal originating from at least part of the vehicle, the method comprising:
- a constitution of a learning base (S1) comprising vibration signals originating from said part of the vehicle, this learning base thus being associated with said vehicle,
- a pickup of a vibratory signal (S2) from a homologous part of a running vehicle,
- an implementation of artificial intelligence to compare (S3) the vibratory signal picked up with the vibration signals of the learning base, and
- as a function of a result of the comparison, a determination (S5) of the current vehicle as corresponding or not to the vehicle (VEHi) associated with the learning base,
process in which:
the vibration signals from the learning base are acquired by said vibratory signal pickup and stored (S6) in the learning base in correspondence of a pickup instant, and
- each time a new vibratory signal is picked up on a current vehicle, if the current vehicle is identified as being the vehicle associated with the learning base, the learning base is updated (S7) by replacing the oldest vibration signal in the learning base by the new vibration signal picked up.
A method according to claim 1, wherein, to identify a particular vehicle (VEHi) among a plurality of vehicles:
a learning base (APPRi) is provided for each vehicle of said plurality of vehicles,
- the capture of a new vibratory signal from a part of the current vehicle is followed by a comparison of the new signal captured at each of the learning bases (S4) with an estimate of a comparison score for each base of learning, and
- the current vehicle is identified (S25) as being the vehicle associated with the learning base whose comparison score (SCOi) is the highest (MAX), the process continuing with the replacement (S26) of the vibration signal the oldest of the learning base with the highest score by the new signal picked up.
Method according to one of the preceding claims, in which the part of the vehicle from which the vibratory signal originates is an engine of the vehicle, and in which the learning base comprises vibration signals characteristic of operating phases of the engine, said phases being marked in the signals of the learning base and the artificial intelligence using marks thus each associated with one of said phases.
A method according to claim 3, wherein said phases include at least one acceleration, one neutral, and one steady state of the engine.
Method according to one of the preceding claims, in which the artificial intelligence uses an artificial neural network with deep learning of the learning base.
Method according to one of the preceding claims, in which the picked up vibratory signal is acquired by a sensor (CAP) of a mobile device (LUN; SM) available to a user of the vehicle, and able to communicate signal data captured at a remote server (SER) storing the learning base (APPRi).
A method according to claim 6, wherein the mobile device (SM) includes an internal battery (BATT) and a connection module (MODC) to a mains power supply (SECT) for recharging the internal battery, and wherein the mobile device is configured to condition the data communication of the signal received to the remote server when the internal battery is recharging.
Method according to one of the preceding claims, in which the picked up vibratory signal is acquired by an acceleration sensor intended to be worn by a predetermined user of the vehicle, the method further comprising an implementation of artificial intelligence for:
- distinguish, in the signals acquired by the sensor, the vibratory signals of the part of the vehicle, gestural signals of the user,
- classify the learning base by distinguishing the vibratory signals and the gestural signals,
- And, in addition to determining the current vehicle as or not corresponding to the vehicle associated with the learning base, determining whether or not there is a correspondence between the user from whom the gestural signals in the picked up signal originate and the user from whom the gestural signals from the learning base.
Method according to Claim 8, in which the acceleration sensor is arranged on a frame of a pair of spectacles (LUN) intended to be worn by the user from whom the gestural signals of the learning base originate, the sensor being capable of being worn by the user. in communicating with a mobile terminal (SM) connected to said learning base.
Method according to one of Claims 8 and 9, further comprising triggering of an alert signal in the event of a mismatch between the user from whom the gestural signals in the picked up signal originate and the user from which originate. the gestural signals of the learning base.
Method according to one of the preceding claims, further comprising:
- an establishment, from the learning base, of a vibration model of said part of the vehicle,
- after picking up the vibratory signal from said part of the vehicle, an estimate of a difference between the vibration model and the picked up vibratory signal, and
- If said estimate is greater than a threshold, triggering of a warning signal of a failure of said part of the vehicle.
A method according to claim 11, further comprising:
- a storage of a plurality of vibration models of homologous parts of vehicles, said homologous parts having respective pre-identified failures, and
- if said estimate of the difference between the model and the signal picked up is greater than the threshold, a comparison of the signal picked up with each of the vibration models of homologous parts with respective failures, to identify the failure of said part of the vehicle from which the vibratory signal picked up.
Computer program comprising instructions for implementing the method according to one of claims 1 to 12, when said instructions are executed by a processor of a processing circuit.
System for implementing the method according to one of claims 1 to 12, comprising at least:
- a memory (MEM) to store the learning base,
- a sensor (CAP) of said vibratory signals, and
- a processing circuit (CT) implementing said artificial intelligence,
the processing circuit being configured to store in the learning base the captured vibration signals, corresponding to a pickup instant, and for each picking up of a new vibratory signal on a running vehicle, in the event of identification of the current vehicle as being the vehicle associated with the learning base, updating the learning base by replacing the oldest vibration signal of the learning base with the new vibratory signal picked up.
.