WO2024028102A1 - Method and system for detecting presence of life in a vehicle - Google Patents

Abstract

The invention relates to a method of detecting presence of life in a vehicle. The method comprises operating at least one radar sensor device (4) mounted in the vehicle to monitor a passenger compartment (1) of the vehicle, the operating comprising transmitting a radar signal towards the passenger compartment of the vehicle and receiving a portion of the transmitted radar signal reflected by a current scenario under a current event in the passenger compartment (1). The received radar signal is pre-processed to obtain at least one parameter characterizing the current scenario under the current event. Then, a weight assigned to the at least one parameter is obtained, wherein the weight is based on a statistical profile for the at least parameter indicating a probability of presence of life in the current scenario under the current event. Based on the obtained weight presence of life in the current scenario is determined.

Description

METHOD AND SYSTEM FOR DETECTING

PRESENCE OF LIFE IN A VEHICLE

The present invention relates to the field of in-vehicle sensor systems. Specifically, the invention relates to a method and system for detecting presence of life in a vehicle by means of at least one radar sensor device.

Modern vehicles, such as cars, are often provided with a variety of comfort, assistant and safety features and may include interior monitoring systems. Such systems may be provided for detecting whether there is a person located on a seat or not. It has been proposed in the art to use radar technology for seat occupant detection systems. This technology may also be used for detecting whether a person, in particular a child or baby, is accidentally left behind in the vehicle, which may be crucial for saving lives, e.g., if the vehicle heats up in the sun. Generally, radar sensor technology may be used to detect presence of life in a vehicle, which may also include, e.g., pets.

The system may output some type of signal or alarm if presence of life is detected when the vehicle is left and locked to inform a user, in particular the vehicle’s driver about a child or pet left behind in the vehicle. This alarm is usually sounded within seconds from the closing and locking of the doors. For example, the signal or alarm can be the vehicle honking loudly, or a signal can be sent to a driver’s smart phone that will attract the driver’s attention. Because of the high urgency, the driver must be alerted within seconds or at the most, a few minutes after walking away from the parked vehicle.

In order to operate reliably and safely, such technology typically would not tolerate any false negative results, i.e., scenarios in which an actually present life is erroneously not detected and no alert is output. In particular, a respective algorithm must be sensitive enough to identify, e.g., a baby sleeping, especially in a baby seat with blankets covering the baby, a sleeping toddler or small pet, or a child or adult who is almost still (e.g., because there are episodes of stopped breathing, which may occur in case of sleep apnea) or is immobile. However, this may possibly cause a high rate of false positive results if the system is oversensitive. While this is more acceptable than false negative results, too many false alerts may be annoying for a user and may even cause a user ignore warnings. False alerts may be caused by other moving objects inside the vehicle, particularly when the vehicle is parked and forced to shake, e.g., by heavy wind or a truck passing nearby. Such moving object may be for instance a filled water bottle or a piece of clothing on a coat hanger. It is an object of the present invention to provide an improved approach for detecting the presence of life in a vehicle. Specifically, it is desirable to improve accuracy and reliability of life presence detection in a vehicle, and specifically to reduce the number of false alerts.

A solution to this problem is provided by the teaching of the independent claims. Various preferred embodiments of the present invention are provided by the teachings of the dependent claims.

A first aspect of the invention is directed to a, particularly computer-implemented, method of detecting presence of life in a vehicle. The method comprises operating at least one radar sensor device mounted in the vehicle to monitor a passenger compartment of the vehicle, the operating comprising transmitting a radar signal towards the passenger compartment of the vehicle and receiving a portion of the transmitted radar signal reflected by a current scenario under a current event in the passenger compartment. The received radar signal is pre-processed to obtain at least one parameter characterizing the current scenario under the current event. A weight assigned to the at least one parameter is obtained, wherein the weight is based on a statistical profile for the at least parameter indicating a probability of presence of life in the current scenario under the current event. Based on the obtained weight, presence of life in the current scenario is determined. If presence of life is determined, a warning signal may be output. Presence of life may be detected if the weight is above a predetermined threshold.

Accordingly, the method may be considered an improved method for detecting presence of life (life presence detection, “LPD”) in a vehicle. The method is particularly advantageous for reducing the number of false alerts. By employing an empiric approach a logic can be implemented that uses weights in a sophisticated manner to evaluate the signal of the radar system. In other words, the usage of statistical profiles instead of heuristic methods simplifies the method of the present invention. There is no need for complicated machine learning that requires larger computational effort and expensive hardware. This leads to a reliable LPD function by means of which, on the one hand, presence of life can be detected with high accuracy, while at the same time false alerts can be avoided or at least reduced. This may significantly improve the safety, particularly for children, babies or pets unintentionally left behind in a vehicle. The statistical profiles may further be used to classify targets, e.g., child, adult, pets, water bottle and other common inanimate items in the vehicle to further improve the method. The term “life presence detection”, as used herein, refers particularly to detection of a living human or living animal, and may be seen particularly in contrast to detection of moving or static objects. Current technology may be configured to detect even small changes inside a vehicle to detect life, e.g., movements of a passenger’s breast caused by breathing, which may include detection of breathing patterns. This may be valid for adult passengers as well as children or babies, and also pets. Thus, the term “life presence detection” (LPD) as used herein may specifically include “child presence detection” (CPD). The LPD function may be typically activated after the ignition of the vehicle is turned off and the doors are closed. The term “LPD function” thus means an activated mode of life presence detection. This function or mode is typically not active, e.g., during driving. The LPD function may be activated for a certain period of time after the vehicle has been parked (e.g., a couple of minutes). Possibly, the LPD function may be switched off once presence of life has been detected.

Terms including “radar”, “radar sensor”, “radar sensor device”, “radar sensor system”, or the like, as used herein, refer particularly to common understanding of a radar or radar system. More specifically, the “radar” can be one radar solution or multiple radar solution in the vehicle’s cabin. Hereinafter, the nomenclature “radar” or “radar system” can encompass either one radar or multiple radar solutions. A one radar solution can be a 3D- or a 4D-radar. 3D refers to three parameters, namely range, velocity, azimuth or elevation angle, while 4D refers to four parameters, namely range, velocity, azimuth and elevation angle. A 3D-radar is typically used to illuminate only one row of the seats while a 4D-radar can illuminate multiple rows, depending on its field-of-view. One unit of this type of radar may be sufficient to output a decision of the LPD function. A multiple radar solution may be made up of multiple 2D-radars. Each 2D-radar can only estimate range and velocity. They tend to be smaller, less complex and cheaper than a 3D- or 4D-radar. Multiple of such 2D-radars may be placed all around the vehicle’s cabin to avoid blind spots. There is a minimum of two 2D- radars if the locations of the moving objects in the car are to be estimated via trilateration techniques. The trilateration operation will be done in a central computing unit. With such a multi-static setup, the decision of the LPD function will most likely come from the computation in the central computing unit. At the beginning of the operation, the radar transmits signals to its surroundings. The reflected radar signals are then processed in a pre-processing unit.

The term “scenario”, as used herein, refers particularly to a configuration of a vehicle passenger compartment including objects in the passenger compartment. Such objects can be living or non-living targets. For example, a current scenario could be a child sitting on a certain seat of the vehicle. The term “event”, as used herein, refers particularly to actions in a certain scenario that may vary over time. Such event could be, e.g., shaking of the vehicle, doors opening or closing etc. Particularly, an event does not change the scenario.

The term “vehicle”, as used herein, refers particularly to a car, including any type of motor vehicles, hybrid electric vehicles and battery electric vehicles as well as other vehicles like trucks, vans, or busses.

If applicable, the terms “first”, “second”, “third” and the like in the description and in the claims, are used for distinguishing between similar elements and not necessarily for describing a sequential or chronological order. It is to be understood that the terms so used are interchangeable under appropriate circumstances and that the embodiments of the invention described herein are capable of operation in other sequences than described or illustrated herein.

Where the term “comprising” or “including” is used in the present description and claims, it does not exclude other elements or steps. Where an indefinite or definite article is used when referring to a singular noun e.g., “a” or “an”, “the”, this includes a plural of that noun unless something else is specifically stated.

Further, unless expressly stated to the contrary, “or” refers to an inclusive or and not to an exclusive or. For example, a condition A or B is satisfied by any one of the following: A is true (or present) and B is false (or not present), A is false (or not present) and B is true (or present), and both A and B are true (or present).

In the following, preferred embodiments of the method are described, which can be arbitrarily combined with each other or with other aspects of the present invention, unless such combination is explicitly excluded or technically impossible.

In some embodiments, the statistical profile includes a table for the at least one parameter for the current scenario under various events. Different events may affect the measurement result of the radar sensor device. Thus, the statistical profile takes into account various events for a certain scenario.

In some related embodiments, a weight is assigned to the at least one parameter for each of the various events. In this way, the method can reliably determine presence of life for different events but the same scenario. In some embodiments, the statistical profile is based on measurements obtained by operating at least one radar sensor device in a reference scenario with a single target in the vehicle passenger compartment under the various events. Reference scenarios having only a single target may represent easy cases of scenarios, e.g., where only one child is sitting in the car with no other moving objects around.

In some embodiments, the statistical profile is based on measurements obtained by operating at least one radar sensor device in a reference scenario with multiple targets in the vehicle passenger compartment under the various events. Reference scenarios with multiple targets, e.g., a child on one seat and a bottle of water on the floor, may provide more complex scenarios that may further improve the results of the LPD method. Statistical profiles of single target scenarios and multiple target scenarios may be combined to further improve the accuracy.

In some embodiments, a plurality of parameters characterizing the current scenario under the current event are obtained, wherein the weight is assigned to a combination of at least some of the plurality of parameters. Having more than one parameter along with combined weights may further improve accuracy of the method as a certain scenario may have a certain combination of characterizing parameters that differs from the combination of parameters for another scenario.

In some embodiments, the at least one parameter comprises a radar parameter that comprises at least one of a Doppler value, a signal-to-noise ratio, a target range, a target location, and a target frequency motion. It will be appreciated that any common radar parameter may be employed.

In some embodiments, the statistical profile further includes a table for an output parameter of an accelerometer for the current scenario under various events. It may be preferable to use data from an accelerometer because such sensor data may deliver reliable results whether the vehicle is shaking or not.

In some embodiments, the method further comprises comparing the obtained weight with history data of shaking events of the vehicle, wherein presence of life is determined based on a result of the comparison. Cross-checking the weighted logic with historical data may improve accuracy of the method. In some embodiments, the method further comprises comparing the obtained weight with history data of actual detection of presence of life in the vehicle, wherein presence of life is determined based on a result of the comparison. Cross-checking the weighted logic with historical data may improve accuracy of the method.

In some embodiments, the method further comprises determining a likelihood of presence of life in the current scenario based on the obtained weight. In other words, rather than outputting a yes/no decision, a continuous probability from 0 to 1 may be output. This may be preferable for a user to let them finally decide based on the probability whether a child is left behind or not. It may also be envisioned to output different levels of warning based on the determined probability.

In some embodiment, the statistical profile is further dependent on a type of vehicle. This may further improve accuracy of the method as the radar results may depend on the type of vehicle, e.g., size and shape of the passenger compartment, number and positions of seats, etc. Furthermore, using vehicle type dependent statistical profiles may allow for easy adaptation to vehicles with different center of gravity or damping factors.

A second aspect of the present invention is directed to a data processing system being configured to perform the method of the first aspect. The data processing system might specifically be configured by means of one or more computer programs to perform the method of the first aspect. In addition, or alternatively, the configuration may be implemented, in whole or in parts by respective hardware.

A third aspect pf the present invention is directed to a radar sensor system, comprising a data processing system according to the second aspect and at least one radar sensor device configured to be mounted in a vehicle to monitor a passenger compartment of the vehicle. It will be appreciated that the radar sensor device may specifically include at least one radar transmitting device for transmitting radar waves towards the vehicle passenger compartment and at least one radar receiving device for receiving reflected radar waves. The radar sensor device may be configured for transmitting radar waves towards such regions in the vehicle passenger compartment in which potential passengers can be expected from positions of seats within the vehicle passenger compartment. Of course, it may also be envisioned to monitor the entire volume of the vehicle passenger compartment, e.g., to be able to detect also pets on the floor. A fourth aspect of the present invention is directed to a computer program or a computer program product, comprising instructions, which when executed on a data processing system according to the second aspect of the invention cause the system to perform the method according to the first aspect of the invention.

The computer program (product) may in particular be implemented in the form of a data carrier on which one or more programs for performing the method are stored. Preferably, this is a data carrier, such as a CD, a DVD or other optical medium, or a flash memory module. This may be advantageous, if the computer program product is meant to be traded as an individual product independent from the processor platform on which the one or more programs are to be executed. In another implementation, the computer program product is provided as a file on a data processing unit, in particular on a server, and can be downloaded via a data connection, e.g., the internet or a dedicated data connection, such as a proprietary or local area network.

The system of the second aspect may accordingly have a program memory in which the computer program is stored. Alternatively, the system may also be set up to access a computer program available externally, for example on one or more servers or other data processing units, via a communication link, in particular to exchange with it data used during the course of the execution of the computer program or representing outputs of the computer program.

The explanations, embodiments and advantages described above in connection with the method of the first aspect similarly apply to the other aspects of the invention.

Further advantages, features and applications of the present invention are provided in the following detailed description and the appended drawings, wherein:

Fig. 1 schematically illustrates one row of a vehicle passenger compartment with one seat occupied;

Fig. 2 schematically illustrates two rows of a vehicle passenger compartment with one seat occupied and an additional item;

Fig. 3 schematically illustrates tables of statistical profile patterns for different scenarios for various events; Fig. 4 schematically illustrates an exemplary embodiment of a method to build weights for profiles under various events; and

Fig. 5 schematically illustrates an exemplary embodiment of a method of detecting the presence of life.

Fig. 1 schematically illustrates a passenger compartment 1 of a vehicle with an example of a single-target profile measurement arrangement of a child 2 sitting on a seat 23. In this example, the child 2 is sitting on the right rear seat 23 of the three seats 21 , 22, 23 of the rear bench of the vehicle. The measurement is carried out by means of a radar sensor device 4 directed to the rear bench of the vehicle. An accelerometer 5 may be provided to further verify a shaking event of the vehicle as will be explained in more detail below. The accelerometer chip 5 may be located on the radar sensor device 4.

Fig. 2 analogously schematically illustrates another profile measurement arrangement. In this case, a multiple-target profile of a water bottle 3 on the floor of the passenger compartment 1 of the vehicle with a child 2 on the right rear seat 23 is determined. Fig. 2 further illustrates the driver seat 1 1 and the co-driver seat 12 for the sake of completeness.

The measured single-target profile and multiple-target profile result in different parameters patterns as will be set forth below. First, general terms for better understanding of the invention will be explained in detail. In the following description, it may be distinguished between “normal cases” and “corner cases” that may occur during operation of the life presence detection (LPD) function.

The term “normal case”, as used herein, refers particularly to cases, in which a vehicle is parked in a normal place such as a parking garage or a designated open parking area. It may be assumed that there are little to no external disturbances that cause the car to shake such as heavy trucks driving by. Typically, once the parked vehicle’s ignition is turned off and the doors are closed, the LPD function turns on. In such a case, which is likely to happen most of the time, the LPD function may be able to detect the presence of child, pet or person with almost 100% true positive.

The term “corner case”, as used herein, refers particularly to the cases that are not “normal cases”. Such a “corner case” or scenario is hereinafter said to lead to a “shaking event” of the parked vehicle. This may be the case, e.g., when the vehicle is parked on a busy street. In that street, there may be possibilities of external disturbances that can cause the parked vehicle to shake. This may tend to happen when a car or truck passes the parked vehicle. The difference in air pressure causes air displacement, which will then shake the parked vehicle.

A “corner case” or “shaking event” may have consequences for the LPD function, which are addressed by the present invention. For instance, when a shaking event happens to a parked vehicle, a radar sensor system may tend to misidentify the detected signals.

A false positive alert may happen when there are loose parts or items in the vehicle that go into a swinging motion such as oscillating and vibrating. Examples for most common items exhibiting such swinging motions are water in a bottle (flask, milk bottle, sippy cup, etc.), items on coat hangers hanging from the hand grip above the door and toys. The swinging motion may resemble a human’s breathing pattern. Hence the LPD function on a radar might not be able to distinguish between a real human and an inanimate object. If the swinging motion happens to be in the range of a human’s breathing rate, it will show up as life is detected.

A false negative may also happen in almost the same way. If the shaking event causes the whole care as well as the inanimate objects in the car to shake with a greater intensity than a human’s rate of breathing, the LPD may deduce mixed signals that can lead to a no life is detected decision. While it is true that the oscillation of water in a bottle, or a coat hanging on a coat hanger in the car will decay over time then stop but there is a chance that it will not decay fast enough (within the time limit to activate the alarm) for such a time critical application. The present invention particularly focuses on mitigating false alerts in the corner cases with an LPD algorithm that uses radar.

Hereinafter, the proposed LPD method may implement a logic that may be referred to as “Weighted Decision Logic”. As will be explained in the context of the following examples, most common outputs (parameters) from a generic radar together with the Weighted Decision Logic LPD algorithm may be used to mitigate false alerts. It has been found that with appropriate parameter tuning, the false alerts under realistic circumstances can be mitigated up to 99.99% of the time. Complicated machine learning can be avoided for cost reduction in hardware and R&D efforts. As explained above, an optional accelerometer chip 5 can also be used. With an accelerometer 5 on board the radar sensor device 4, there will be one more information source about the shaking events on the parked vehicle. This information can be used to refine the LPD’s decision logic for better decision accuracy and to make faster decisions. The proposed Weighted Decision Logic LPD algorithm can be used with any radar using any type of waveforms so long as certain common parameters can be seen at the output of the radar’s pre-processing unit. This will be explained in the following sections. It is also imperative that the radar’s parameters have been configured to suit the characteristics of its surroundings. For example, the range and velocity resolution must be small enough to capture detections in a small space of the vehicle’s cabin. Furthermore, sufficient antenna elements should be used for a sufficiently fine angular resolution. In the exemplary embodiment described herein, a radar operating at 60GHz with 3Tx-4Rx was used to validate the LPD algorithm.

Referring now to Fig. 3 exemplary tables (a), (b) and (c) of profile parameters with parameters 1 ... N are provided for different scenarios in the vehicle passenger compartment 1 under various events. The parameters refer to common radar parameters (outputs), and may comprise at least one of the parameters A, B, C, D, and E set forth below.

Common outputs of the pre-processing unit may be (A) Power or Signal-to-Noise Ratio (SNR) of the multiple targets. This can include power per range/Doppler/angular bin or average power. Further parameters may be (B) the range (distance from the radar to the targets) or (C) Doppler or velocity (of targets relative to the radar). Apart from that, (D) locations of the multiple targets can be considered, e.g., in terms of azimuth and/or elevation angles, spherical coordinates in [R,e,<p] axes with R= radius, 9,<p being the azimuth and elevation angles, or Cartesian coordinates in [x,y,z] axes. Further, (E) frequency motions of the targets can be considered. This can be divided into normal breathing rate frequency - the range includes the breathing rates for babies, children, adults and common household pets; high or low frequency motions - most likely to be inanimate objects like toys or water in bottles; random motions - where no regular pattern of frequency can be identified; harmonics - frequencies that are induced by the fundamental or true frequency (a harmonic is always n times the fundamental frequency, whereby n=2,3,4,....). A harmonic frequency is most likely a result of mechanical vibrations instead of normal breathing. The aforementioned parameters A, B, C, D, E may be available for each data bin i.e., per range/Doppler/angular bin.

It is common to use data buffers to accumulate data over time. The data can be the raw or processed data of various processing stages. There may be at least two data buffers, which store the data over different time lengths. Shorter buffers store less data and are used to generate fast results. These results can be erroneous because of the short observation time. Longer buffers store more data within a longer time and are used to generate results that are more accurate. This is because an observation over a longer time will yield a more stable picture of the environment. The data buffer sizes may be designed based on the radar waveforms and timings used. This is to ensure that a long enough observation time is kept for the shorter buffers to glean meaningful responses in frequency and power. The size of the longer buffers should not exceed the limit, where the state of the objects and humans in the cabin could have changed significantly. The aforementioned parameters A, B, C, D, E may be available for each data bin i.e., per range/Doppler/angular bin and per buffer.

The profile patterns shown in Fig. 3a, 3b and 3c for each event are made up of exemplary parameter combinations. The parameters denoted as Param 1 , 2, 3, 4, ..., N can be or can comprise the common radar parameters as given above in A, B, C, D, E as well as any external parameters such as accelerometer, vehicle host data, camera data, etc. The example profile patterns show different events along the timeline 0 ... 19, including door open and close (time 1 - 2), shaking event (time 6 - 12), stationary (no shaking) (time 16 - 19). After a start at time 0, there exist some transition times from one state to another (time 3 - 5 and time 13 - 15). In the respective columns of time there are marked with an “x” in the respective parameter line if the radar sensor device 4 has an output signal in this respect.

Fig. 3a shows a statistical profile relating to a child present in a vehicle as illustrated in Fig 1 . Fig. 3b shows a profile of a water bottle only (cf. water bottle as shown in Fig. 2.). Fig. 3c shows a profile of the water bottle and child illustrated Fig. 2. The profile seen here is usually a mix of the two other aforementioned profiles. This is used as a confirmation and tuning of the weighted probabilities.

An exemplary method 100 to build weights is now described with reference to Fig. 4. This methodology uses the aforementioned obtained profile patterns of Fig. 3. Start and end are indicated at S1 1 and S16, respectively.

First, measurement campaigns (S12) are done to obtain statistical profiles of single targets such as adult, child, pet, water in bottles, toys, and other common in-car inanimate items as illustrated in Fig. 1. These measurements are done for various events, including both shaking and no-shaking events to the vehicle. Each profile will be made of ‘pattern’ (see Fig. 3a, Fig. 3b), which may be a combination of different parameters of A, B, C, D, E from each data buffer. Next, statistical profiles of common multiple targets are determined (S13). Examples are a child with milk bottle/ toy/ pet, coat hangers on the hand rail, water bottles on the floor or non-cushioned spaces (see Fig. 2). In the same way, the measurements are done for various events, including both shaking and no-shaking events to the vehicle (cf. Fig. 3c).

Tables like those shown at Fig. 3a, 3b and 3c are then build to map statistical probabilities (S14). They are then weighted based on their statistical probability of happening across the multiple profiles (S15). In other words, weights are assigned to the parameters or combinations of parameters for the various events.

Should the radar possess outputs in addition to the parameters of A, B, C, D, E from each data buffer, they can also be taken into account to build the profiles. Should there be additional redundancy information besides the parameters and accelerometer data mentioned above, they can able be taken into account to build the profiles. Examples are camera data, smart access sensors in the cabin that registers movement, vehicle host data.

For example from Fig. 3a and 3b, Param 3 and 4 show up in both single-target profiles during a shaking event. A check in the multiple-target profile in Fig. 3c confirms this mixed of Param 3 and Param 4. The weighted logic in this case may look like this in the algorithm: if (Param 3 && Param 4), then life_detection = 0.5. Or said in words, if Param 3 and Param 4 appear together, the probability of a life detection due to a human is 50%.

When the parameters or combination of parameters have been identified with a statistical probability assigned for all events based on the single-target profiles as explained by way of the example above, the methodology ends (S16). The next stage may be the top level Weighted Decision Logic algorithm described below with reference to Fig. 5.

Referring now to Fig. 5, a method 200 of life presence detection using an exemplary weighted decision logic LPD algorithm implementation is explained starting at S21 and ending at S28. In this top level algorithm, the weighted logic is crossed checked with historical data to deliver the final decision of whether life is detected or not. First, radar preprocessing with outputs of Param 1 , 2, 3, ..., N is performed (S22). This may be understood as the execution of the LPD function after the doors of the vehicle have been closed. The following steps are carried out to make a decision whether life is present in the vehicle or not. Based on the assigned weights as described above, a decision in the form of a likelihood between 0 and 1 will be output (S23). The likelihood of may be, e.g., 0.5. The historical data on the statistical probability of life detection is also accumulated over a period of time. If an accelerometer is used, historical data on the detection of shaking events are also recorded. In a step 24 it may then be determined whether the history of a shaking event is above a certain threshold, and if yes, a final decision with a likelihood between 0 and 1 can be output (S26). For instance, it could be more likely in this case that there is no life present but this was merely a shaking event. Continuing with step S25 however, it may then be determined whether the history of life detected is above a certain threshold. If yes, a final decision with a likelihood between 0 and 1 can be output (S27). For instance, an alarm signal can be output that there has been detected the presence of life. It is advantageous if the final logic depends on the historical data since the confidence level will be higher.

Usage of continuous probability from 0 to 1 instead of merely ‘0’ or ‘T especially in the first few seconds when the confidence of the output is low (due to no historical data available) may be advantageous as it allows to assign different warning levels based on the continuous probability. This could be used to give different alarms to the user (e.g., loud honk or only a light signal).

The above methodology can be used to classify child, adult, pets, water bottle and other common inanimate items in the car. An easy adaptation may be possible to cars with different center of gravity or damping factors. While it will be appreciated that machine learning can be used in this method, there is no absolute need for complicated machine learning that requires larger computational effort and expensive hardware.

While above at least one exemplary embodiment of the present invention has been described, it has to be noted that a great number of variations thereto exists. Furthermore, it is appreciated that the described exemplary embodiments only illustrate non-limiting examples of how the present invention can be implemented and that it is not intended to limit the scope, the application or the configuration of the herein-described apparatuses and methods. Rather, the preceding description will provide the person skilled in the art with constructions for implementing at least one exemplary embodiment of the invention, wherein it has to be understood that various changes of functionality and the arrangement of the elements of the exemplary embodiment can be made, without deviating from the subject-matter defined by the appended claims and their legal equivalents. LIST OF REFERENCE SIGNS passenger compartment of vehicle child bottle radar sensor device accelerometer driver seat co-driver seat left rear seat middle rear seat right rear seat

Claims

CLAIMS A method of detecting presence of life in a vehicle, the method comprising: operating at least one radar sensor device (4) mounted in the vehicle to monitor a passenger compartment (1 ) of the vehicle, the operating comprising transmitting a radar signal towards the passenger compartment of the vehicle and receiving a portion of the transmitted radar signal reflected by a current scenario under a current event in the passenger compartment (1 ); pre-processing the received radar signal to obtain at least one parameter characterizing the current scenario under the current event; obtaining a weight assigned to the at least one parameter, wherein the weight is based on a statistical profile for the at least parameter indicating a probability of presence of life in the current scenario under the current event; and determining presence of life in the current scenario based on the obtained weight. The method of claim 1 , wherein the statistical profile includes a table for the at least one parameter for the current scenario under various events. The method of claim 2, wherein a weight is assigned to the at least one parameter for each of the various events. The method of any one of the preceding claims, wherein the statistical profile is based on measurements obtained by operating at least one radar sensor device (4) in a reference scenario with a single target in the vehicle passenger compartment (1 ) under the various events. The method of any one of the preceding claims, wherein the statistical profile is based on measurements obtained by operating at least one radar sensor device (4) in a reference scenario with multiple targets in the vehicle passenger compartment (1 ) under the various events. The method of any one of the preceding claims, wherein a plurality of parameters characterizing the current scenario under the current event are obtained, wherein the weight is assigned to a combination of at least some of the plurality of parameters.

. The method of any one of the preceding claims, wherein the at least one parameter comprises a radar parameter that comprises at least one of a Doppler value, a signal-to-noise ratio, a target range, a target location, and a target frequency motion. . The method of any one of the preceding claims, wherein the statistical profile further includes a table for an output parameter of an accelerometer for the current scenario under various events. . The method of any one of the preceding claims, further comprising comparing the obtained weight with history data of shaking events of the vehicle, wherein presence of life is determined based on a result of the comparison. 0. The method of any one of the preceding claims, further comprising comparing the obtained weight with history data of actual detection of presence of life in the vehicle, wherein presence of life is determined based on a result of the comparison. 1. The method of any one of the preceding claims, wherein further comprising determining a likelihood of presence of life in the current scenario based on the obtained weight. 2. The method of any one of the preceding claims, wherein the statistical profile is further dependent on a type of vehicle. 3. A data processing system, the system being configured to perform the method of any one of claims 1 to 12. . A radar sensor system, comprising a data processing system according to claim 13 and at least one radar sensor device (4) configured to be mounted in a vehicle to monitor a passenger compartment (1 ) of the vehicle. 5. A computer program or a computer program product, comprising instructions, which when executed on one or more processors of a system according to claim 14 cause the system to perform the method according to any one of claims 1 to 13.