Microphone-based event detection device for a motor vehicle, and method for detecting at least one specific event

The microphone-based event detection device addresses the challenge of reliable event detection in vehicle monitoring systems by using sound sensors and FD-SOI technology to analyze sound characteristics, achieving efficient and continuous monitoring with reduced energy consumption.

WO2024240537A1PCT designated stage expired Publication Date: 2024-11-28AUDI AG +1
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2024/063165
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-05-23
Filing Date
2024-05-14
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Existing vehicle monitoring systems face challenges in reliably detecting specific events, such as theft attempts, in an inactive state without excessive energy consumption, leading to frequent false alarms and limited monitoring capabilities.

Method used

A microphone-based event detection device using sound sensors and processing units, including FD-SOI technology, to detect sound-generating events by analyzing sound characteristics and distinguishing between different events, with energy-saving modes to enable permanent monitoring.

Benefits of technology

The system provides reliable and differentiated event detection with reduced false triggers and low energy consumption, allowing for continuous monitoring of vehicles, even in inactive states, and enabling various applications like voice control and component monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2024063165_28112024_PF_FP_ABST
    Figure EP2024063165_28112024_PF_FP_ABST
Patent Text Reader

Abstract

The invention relates to an event detection device (12) for a motor vehicle (10), said device being designed to detect at least one specific event (E1, E2) in an inactive state (ZI) of the motor vehicle (10). The event detection device (12) comprises a processing device (18) and at least one acoustic sensor (14, 16) which is designed to detect sound (S1, S2) transmitted via air, wherein the event detection device (12) is designed to detect an event (E1, E2) which produces sound (S1, S2) as the event (E1, E2), detect the sound (S1, S2) produced by the event (E1, E2) as a sound signal (S1, S2) by means of the acoustic sensor, ascertain a sound characteristic (C1, C2) from the detected sound signal (S1, S2), and detect the event (E1, E2) on the basis of the sound characteristic.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Microphone-based event detection device for a motor vehicle and method for detecting at least one specific event

[0002] DESCRIPTION:

[0003] The invention relates to an event detection device for a motor vehicle, wherein the event detection device is designed to detect at least one specific event in an inactive state of the motor vehicle. Furthermore, the invention also relates to a method for detecting an event.

[0004] An alarm system, for example, can be considered an event detection device designed to detect at least one specific event when a motor vehicle is inactive. Such alarm systems are typically used to prevent a possible theft of the motor vehicle. Various vehicle sensors can be used to detect such a possible theft attempt, for example, door contact switches to detect the opening of the doors when the motor vehicle is locked. Inclinometers or position sensors of the motor vehicle can also be used, for example, to detect changes in the vehicle's position.

[0005] In principle, numerous different sensors, such as cameras, ultrasonic sensors or similar, are known from the state of the art, which would in principle enable comprehensive monitoring of a motor vehicle. However, the energy consumption of such systems is particularly problematic when monitoring a motor vehicle when it is in an inactive state. To prevent the 12-volt battery from being over-discharged when the vehicle is inactive and, for example, parked, the vehicle monitoring mechanisms used to date have been limited to relatively basic monitoring of the vehicle, as described above in connection with the alarm system. For example, previous systems cannot reliably distinguish between an actual theft attempt of the vehicle and situations in which there is no theft attempt at all. This leads to frequent false alarms.Furthermore, it would be desirable to be able to not only detect a theft attempt when the vehicle is inactive, but also to monitor it for the possible occurrence of other events.

[0006] The object of the present invention is therefore to provide an event detection device for a motor vehicle and a method that enable more reliable and differentiated detection of a specific event. Furthermore, advantageous developments of the invention aim to provide such a detection option, if possible, as a permanent detection option for the motor vehicle.

[0007] This object is achieved by an event detection device and a method having the features according to the respective independent patent claims. Advantageous embodiments of the invention are the subject of the dependent patent claims, the description, and the figures.

[0008] An event detection device according to the invention for a motor vehicle is designed to detect at least one specific event when the motor vehicle is inactive. The event detection device comprises a processing device and at least one sound sensor configured to detect airborne sound. The event detection device is designed to detect a sound-generating event as the event, to capture the sound generated by the event as a sound signal using the sound sensor, to determine a sound characteristic from the detected sound signal using the processing device, and to detect the specific event depending on the determined sound characteristic.

[0009] The invention is based on the realization that there are numerous different events which, when they occur, generate a characteristic sound signal. This characteristic sound signal can be used to identify and detect such an event accordingly. This advantageously makes it possible not only to detect a specific event, but in particular also to detect numerous different events and differentiate them from one another. Furthermore, even when considering only a single such specific event, it can be detected much more reliably, and false triggering or false detections can be avoided more effectively. Furthermore, the detection of an event based on the sound generated by the event enables particularly energy-saving detection of such an event, as will be explained in more detail later.This energy-saving training option in turn makes it possible to provide continuous monitoring of the motor vehicle by the event detection device.

[0010] The fact that the event detection device is designed to detect such a specific event when the motor vehicle is inactive does not imply that the detection of such events is limited to the inactive state of the motor vehicle. In particular, the event detection device can also be designed to detect such a specific event when the motor vehicle is inactive. However, such a detection option is particularly advantageous when the motor vehicle is inactive, when there is generally no user or driver of the motor vehicle in the vicinity of the motor vehicle, since events occurring in such a situation cannot be directly perceived by the driver or user themselves, who are not even present. In addition, numerous other application possibilities open up here, e.g.Voice control of vehicle components, even if the person speaking is outside the vehicle and the vehicle is in an inactive state.

[0011] The sound generated by such a specific event does not necessarily have to be in the audible frequency range, although this can also be the case. Furthermore, sound can generally be understood to mean sound transmitted via air or a gaseous medium, as well as structure-borne sound, i.e. sound transmitted via a solid body. In other words, the event that can be detected by means of the event detection device can also be an event that generates structure-borne sound, e.g. in addition to a sound signal transmitted via air, with the structure-borne sound then being transmitted, for example, via the vehicle body of the motor vehicle. This advantageously also makes it possible to detect events that generate, for example, vibrations, shock signals or the like. For the purposes of the present invention, however, the sound signal used is preferably one that can be transmitted via air or a gaseous medium.transmitted sound signal is recorded and evaluated, whereby the event generating the sound can optionally also generate, in addition to this sound signal transmitted via air, additional structure-borne sound transmitted via the body, which then, however, does not necessarily have to be recorded and evaluated, and preferably is not recorded and evaluated either.

[0012] The sound sensor used to detect the sound signal is designed at least to detect airborne sound. Optionally, the sound sensor can also be designed to additionally detect a structure-borne sound signal. The sound sensor is preferably designed as a microphone, e.g., as a MEMS (micro-electro-mechanical system) microphone. This is a microphone provided by a micro-electromechanical system. MEMS microphones are miniaturized microphones designed using SMD (surface-mounted device) technology for direct use on electronic circuit boards.

[0013] It is also possible for the event detection device to have several different sound sensors or several similar sound sensors, which can be arranged, for example, at different positions in the motor vehicle, e.g., several microphones and / or a microphone array. These can be coupled to a common processing device. The sound sensors can then be designed as separate structural units, in particular separate from one another and separate from the processing device. Alternatively, each sound sensor can be assigned to its own processing device. Such a sound sensor and the processing device can form a structural unit. The event detection device can then accordingly comprise several such sensor units, each with a sound sensor and an associated processing device.Sensors for detecting airborne sound, such as microphones, generally have a relatively strong directivity in their reception characteristics. Therefore, when using such sound sensors, such as microphones, it is advantageous to provide several such microphones, positioned and aligned at different locations on the vehicle, for example, depending on the event to be detected. However, only one such event detection device with a single microphone can be provided on the vehicle, and placed in a very advantageous position. The microphone can be mounted, for example, on a bumper of the vehicle in order to capture realistic sound data.

[0014] In order to detect the specific event depending on the determined sound characteristic, at least one assigned reference characteristic can be specified for the specific event. This can be stored, for example, in a memory of the event detection device. Such a reference characteristic, as well as the derived sound characteristic, can be a frequency spectrum, for example, as will be explained in more detail later. By comparing the derived sound characteristic with the reference characteristic, the event detection device can determine whether there is a sufficiently high level of similarity or agreement. This can be established, for example, by means of a predetermined limit value for the similarity or agreement. If this is the case, the event can be considered detected, for example. The detection of the event can also depend on further criteria.For example, a subsequent plausibility check can also take place. It is also conceivable that several different assigned reference characteristics are specified for a specific event. This is particularly advantageous for events that can generate different sound signals. To give an example, the event to be detected could be hail. If hail hits the vehicle, this generates sound. However, the sound signal generated can vary depending on the situation and depends, for example, on the intensity of the hail, the size of the hailstones, the surface area of ​​the vehicle exposed to the hail, and so on. For the event "hail", various reference characteristics that take these different situations into account can be stored. This allows the event "hail" to be detected more clearly and reliably.Accordingly, the sound characteristic derived from the detected sound signal can also be compared with multiple reference characteristics assigned to the same event, and the event can be detected based on the result of the comparison. Furthermore, the assigned reference characteristic does not necessarily have to be a fixed, stored reference characteristic, such as a fixed frequency spectrum; rather, the event detection device can also be based on a learning process for identifying the event, for example. This also enables adaptation or modification of the assigned reference characteristic during the learning process.

[0015] In a further very advantageous embodiment of the invention, the event detection device has at least one FD-SOI (fully depleted silicon on insulator) semiconductor component, in particular as part of the processing device of the event detection device, which comprises, for example, an integrated circuit with an FD-SOI substrate.

[0016] FD-SOI is a process technology in which a thin silicon layer, which in the case of a transistor as a semiconductor component, for example, forms the channel of the transistor, is separated from the underlying silicon substrate by an equally thin insulating layer. This technology prevents the flow of electrons from the channel into the substrate and therefore enables significantly more energy-efficient operation of the processing device or the event detection device itself. In principle, any semiconductor components can be manufactured using this FD-SOI technology. In particular, not only individual discrete semiconductor components, such as transistors, can be based on this FD-SOI technology, but also, for example, the chip of the evaluation unit, which provides the integrated circuit, can be provided as an FD-SOI chip. The MEMS microphone can also be manufactured using FD-SOI technology.If, for example, the processing device is designed as an integrated circuit or has such an integrated circuit, this integrated circuit can therefore comprise an FD-SOI substrate. However, even if the processing device is constructed from direct electronic components, e.g. transistors, these can be manufactured using FD-SOI technology. This FD-SOI technology thus advantageously enables extremely energy-efficient operation of the event detection device. This advantageously enables the continuous use of the event detection device even when the motor vehicle is inactive. In particular, this can provide permanent monitoring of the motor vehicle, regardless of the current operating state of the motor vehicle. For example, the event detection device can, e.g.Unless deactivated, for example, by a user or driver, it must be permanently active—that is, 24 hours a day, 7 days a week, all year round, and for several years. This also allows for advantageous application possibilities for this event detection device, even independent of the vehicle's drive type. This can be, in particular, a combustion engine-powered vehicle, a hybrid vehicle, or an electric vehicle.

[0017] In a further very advantageous embodiment of the invention, the event detection device has a sleep mode and an active mode, wherein the event detection device is designed to detect sound signals in sleep mode and to classify them as relevant or irrelevant according to at least one specific relevance criterion, and to switch to active mode if a sound signal is classified as relevant, in particular wherein power consumption by the event detection device is lower in sleep mode than in active mode. This is particularly advantageous because in sleep mode, in which significantly less power is consumed, a certain pre-selection of received sound signals can already take place. Pre-processing and / or more detailed analysis of a detected sound signal only takes place if it has previously been classified or classified as possibly relevant or as relevant in sleep mode.More energy-intensive computing operations that can occur in active mode can thus advantageously be avoided in sleep mode. This allows the overall power consumption of the event detection device to be kept very low in the long term. This is particularly advantageous because a motor vehicle is typically constantly exposed to sound-generating environmental influences. However, most of these environmental influences are not associated with any event to be detected and are therefore irrelevant. Such irrelevant sound signals can thus be filtered out from the outset in sleep mode with extremely low energy consumption. This can increase the energy efficiency of the event detection device's operation.

[0018] In this case, it represents a particularly advantageous embodiment of the invention if the processing device has a basic processing unit, in particular a sound pressure detection unit, wherein the processing device is set up such that the basic processing unit is active in sleep mode, in particular as the only component of the processing device, and classifies the detected sound signal as relevant or not relevant according to the determined relevance criterion, in particular wherein the basic processing unit is designed to provide a wake-up signal depending on a signal level of the detected sound signal.

[0019] The predetermined relevance criterion, based on which sound signals detected in idle mode are classified as relevant or irrelevant, can thus include, for example, that the detected sound signal has a certain minimum signal level, particularly in an unfiltered state and / or with respect to at least one specific frequency range of interest, to which the received signal can be limited by prior frequency filtering, e.g., using one or more filters. The signal level can be detected as sound pressure.The sound pressure detection unit can be designed as a monoband sound pressure detection unit for monoband sound pressure detection, which can determine whether the sound signal as a whole or within a single frequency band has a certain minimum sound pressure or not, or the sound pressure detection unit can be designed as a multiband sound pressure detection unit for multiband sound pressure detection, i.e. for detecting whether a respective, specific, optionally different minimum sound pressure is reached or not in different frequency bands, which can be provided, for example, by different filters. The type of pre-filtering that takes place can also be selected, for example, depending on which specific events are to be detectable by the event detection device.

[0020] Optional pre-filtering in the form of frequency filtering can be implemented electrically very easily and with very little energy consumption. This allows numerous irrelevant sound signals to be filtered out in advance. Checking for the presence of a minimum sound level or minimum sound pressure of the sound signal can also be easily performed using a threshold method, i.e., by comparing it with a predefined threshold. Different thresholds can also be specified for different frequency bands. The detected sound signal can, for example, be classified as relevant if the predefined minimum sound level is reached or exceeded in one or more frequency bands.

[0021] The event detection device can be configured so that only when a sound signal has been classified as relevant is the event detection device awakened, or switches from sleep mode to active mode, in order to perform more complex evaluation and processing of the signal. The preprocessing and classification of the signal as relevant or irrelevant, as well as the wake-up functionality, can be performed using a very simple electrical circuit as part of the processing device and therefore consumes little energy.

[0022] In general, background noise or interference can also be initially filtered out from a sound signal to be analyzed. Suitable background noise reference spectra can also be stored for such background noise or interference, which are typical, for example, for a vehicle operating at a ferry. This allows a more reliable distinction to be made between the wanted signal and the interference signal. This can also be done as part of the pre-filtering and pre-selection performed in the detection device's idle mode and / or in active mode as part of a more in-depth analysis.

[0023] According to a particularly advantageous embodiment of the invention, it is provided that the processing device comprises a plurality of processing modules, wherein each of the processing modules is designed to execute an associated processing task, wherein each of the processing modules has a first idle state and a second active processing state, wherein the event detection device is configured such that the processing modules are in the respective first idle state in the idle mode of the event detection device, and wherein the event detection device is configured such that the associated processing task can be executed by a first of the processing modules, which is in the associated second active processing state, while at least a second of the processing modules is in the associated first idle state. A respective processing module can be a function block.In particular, a respective processing module can be provided in the form of software and / or hardware. The special feature of this embodiment is that each processing module has or can assume its own idle state and active processing state, and e.g. each processing module can be woken from the idle state separately and independently of the other processing modules. If the event detection device is in idle mode, all processing modules are also in their respective idle state. If the event detection device changes to the active state, not all of the processing modules necessarily have to be activated and switch to the active processing state. For example, only the currently required levels, i.e. the currently required processing modules, are activated and the others are put into the respective first idle state or.left in this state and, wherever possible, operated in a power-saving mode, such as clock gating and / or power gating. A particular processing module consumes less energy in its assigned idle state than in its assigned active processing state. This selective or successive awakening of the processing modules or processing levels can therefore result in enormous additional energy savings.

[0024] According to a further particularly advantageous embodiment of the invention, it is provided that the event detection device is set up in such a way that the plurality of processing modules can be woken up in a predetermined order and, when woken up, change from the first inactive idle state to the second active processing state, in particular wherein the event detection device is set up in such a way that a first of the processing modules, which has executed the assigned processing task in the active processing state, provides a wake-up signal depending on a result of the executed processing task, depending on which wake-up signal a next one of the processing modules is woken up according to the predetermined order in order to execute the processing task assigned to it.

[0025] Thus, if the event detection device switches to the active state, not all of the processing modules necessarily have to be activated and switch to the active processing state at the same time. On the contrary, the individual processing modules can be woken up, at least partially, successively or cascaded. The cascading of functions preferably occurs with increasing processing complexity and, associated with it, ever-increasing power consumption. This means that the entire cascade only needs to be woken up in rare cases. Overall, a significant reduction in power consumption can be achieved with appropriate dimensioning of the wake-up parameter thresholds.

[0026] Each processing module can therefore provide a separate processing level that is only activated and requires power (and loss) once a detection has occurred in the previous level. This wakeup cascade enables, for example, particularly low-loss detection of speech and / or other noise-based events.

[0027] The wake-up functionality can be provided by a system controller. The wake-up signal provided by the first processing module can, for example, first be transmitted to this system controller, which then, for example, wakes up the next processing module in the sequence, in particular by transmitting a corresponding wake-up signal to this processing module.

[0028] The processing modules can be implemented on a power-saving microcontroller platform. As explained in more detail later, one of the processing modules can also be provided by an optional microcontroller. The wake-up routine, which can also be referred to as a wake-up routine, is preferably implemented not by a microcontroller, but by a different system controller, namely the system controller mentioned above. This also saves energy. The system controller can be designed particularly simply and be dedicated solely to the task of waking the processing modules.

[0029] The system controller can, for example, be woken up by the wake-up signal provided by the basic processing unit when the detected sound signal has been classified as relevant. The system controller can then initially wake up a first processing module, e.g. for pre-processing of the detected sound signal. Only when the pre-processing is complete is a second processing module woken up to perform a further processing task. If it is determined during one of these processing steps that the detected sound signal cannot be assigned to a specific event to be detected, the wake-up chain can be aborted, i.e. the other processing modules are not woken up at all and the event detection device can return to sleep mode in which only the basic processing unit is active.

[0030] This concept of successive wake-up stages with decreasing false / positive detection rates and inherently increasing energy consumption of the stages leads to a minimized energy consumption of the entire wake-up system, and thus of the entire event detection device.

[0031] In a further advantageous embodiment of the invention, the event detection device in active mode is designed to generate a frequency spectrum of the sound signal classified as relevant, in particular by means of a Fourier transformation of the sound signal, and to compare the generated frequency spectrum, which represents the sound characteristic derived from the detected sound signal, with the reference characteristic provided as a reference frequency spectrum. A frequency spectrum indicates the frequencies contained in the sound signal with their associated amplitudes and phases. Observing a frequency spectrum to identify and detect specific events is particularly advantageous because different events can be detected and identified particularly clearly and thus particularly reliably based on their frequency spectra.This allows for easy differentiation between numerous different events. Furthermore, it allows for a simple functional expansion of the event detection device via software. For example, the functionality of the event detection device can be expanded to include the detection capability of additional new events simply by supplying the event detection device with corresponding reference characteristics in the form of corresponding reference frequency spectra for such additional events via software and / or storing them in the event detection device's memory. It would also be conceivable for a user to be able to select, for example, from a database which events should be detected by the event detection device. These events can then be loaded into the event detection device in the form of their assigned reference characteristics.

[0032] Such a frequency analysis is preferably performed only in the active mode of the event detection device, particularly by one of the processing modules. This enables particularly power-efficient operation of the event detection device, which can be limited to the execution of a few basic functions.

[0033] Therefore, it represents a further advantageous embodiment of the invention if the plurality of processing modules comprise a first processing module which, in the active mode of the event detection device and in its second processing state, is designed to carry out pre-processing of the detected sound signal classified as relevant and to provide a result of the pre-processing, in particular the sound characteristic determined in the process, to at least one second processing module, in particular to several of the processing modules, if these are awakened to carry out their respective processing task, wherein the pre-processing preferably comprises frequency filtering and / or generating a frequency spectrum of the sound signal classified as relevant.

[0034] Such a frequency spectrum can be generated using a Fourier transform. There are various ways to implement this. One example is an FFT (Fast Fourier Transformation). Depending on the application, other variants of frequency spectrum generation may also be provided. A Fourier transform can also be performed using the Goertzel algorithm and / or implemented as a logarithmic Fourier transform and / or used to generate mel frequency cepstral coefficients.

[0035] The Goertzel algorithm is a method from digital signal processing and represents a special form of the discrete Fourier transform. In contrast to the various fast calculation methods for the discrete fast Fourier transform, which always calculate all discrete spectral components in a block, the Goertzel algorithm allows only individual discrete spectral components to be calculated. Its applications lie in the detection of individual frequencies (tone detection) in a signal. This is where the algorithm is particularly efficient.

[0036] Mel Frequency Cepstral Coefficients (MFCC) are used, for example, for automatic speech recognition. They result in a compact representation of the frequency spectrum.

[0037] In this case, i.e., during preprocessing, a reduced calculation accuracy can be achieved through appropriate implementation if the subsequent classification algorithms, which can be implemented by one or more additional processing modules, exhibit sufficient robustness. In addition, the sampling rate, especially for sampling the sound signal to generate the spectrum, can be dynamically adjusted depending on the application.

[0038] During preprocessing, the first processing module can also prefilter the sound signal, particularly by high-pass and / or low-pass and / or band-pass filtering. This has the significant advantage that only the frequencies of interest are considered in advance and, if necessary, examined in the subsequent analysis. This filtering can, in particular, be performed prior to the generation of the spectrum or spectral analysis. In other words, the frequency spectrum can be generated from the filtered sound signal and / or the spectral analysis can be based on the filtered sound signal.

[0039] The results provided during preprocessing, in particular the generated frequency spectrum, can represent the sound characteristics determined from the acquired sound signal. These results provided by the first processing module can also be used by further processing modules, in particular all other processing modules, to perform their respective processing tasks, should the respective processing modules be awakened.

[0040] The additional processing modules can generally be used to implement classifiers for classifying the determined sound characteristics and assigning them to one of several specific events, e.g. based on neural networks. In principle, the event detection device can be designed such that only a single specific event can be detected. During the classification, it can then be determined whether this event occurs or not. The event detection device can also be designed such that several different specific events can be detected. During the classification, it can then be determined whether one of these events occurs or not, and if so, which of the specific events occurs. This will be illustrated again using the following examples, which primarily relate to speech recognition but can be implemented similarly for the detection of other specific events.

[0041] According to a further advantageous embodiment of the invention, it is provided that the plurality of processing modules comprise a second processing module which is designed to carry out speech pause detection, in particular depending on the result of the preprocessing, and / or comprise a third processing module which is designed to carry out word recognition, in particular keyword recognition, in particular depending on the result of the preprocessing, and / or;comprise a fourth processing module designed to carry out speech recognition, in particular depending on the result of the preprocessing, and / or comprise a fifth processing module which represents a microcontroller designed to carry out at least one function, in particular depending on the result of the preprocessing and / or a result of the processing task carried out by the second and / or third and / or fourth processing module;

[0042] As higher recognition levels, which can be implemented by the second and / or third processing module, “Voice Activity Detection” (VAD), ie speech pause detection, and “Key Word Spotting” (KWS), in particular word recognition, e.g. keyword recognition, are part of the concept in this case.

[0043] Speech pause detection (VAD) is a technique used in speech processing to detect the presence or absence of a human voice. Various algorithms can be used for speech pause detection, each with different characteristics and representing a trade-off between latency, sensitivity, precision, and computational effort. Some algorithms also provide further analysis data, for example, whether the speech is voiced, unvoiced, or sustained. Speech pause detection can be independent of the language used. Using speech pause detection, the second processing module can determine whether the detected sound signal is speech or not. If not, the event detection device can return to sleep mode. If it is, the third processing module can be woken up. This can check whether the detected sound signal contains a specific word, e.g.a keyword. If not, the event detection device can return to sleep mode. If it does, the optional fourth processing module can be awakened. This can be used to implement automatic speech recognition. However, even without this fourth processing module, numerous advantageous speech-based applications can be implemented using the event detection device. Eliminating the fourth processing module also enables further energy savings.

[0044] For example, magic commands, i.e., voice commands, in particular key voice commands, can be detected by the event detection device, e.g. even without the fourth processing module. This is made possible, as described above, e.g., by the third processing module. These magic commands are primarily commands or keywords or key phrases for initializing a voice input mode of the motor vehicle. Therefore, within the scope of detecting such a voice command, it is not necessary to correctly understand and interpret the voice commands, which can trigger various functions on the motor vehicle, itself, i.e., by the event detection device, which is why the fourth processing module can be dispensed with in this application. This means thatWith Magic Commands, the detection of an initialization command is particularly important, as it can also be easily recognized based on an associated reference characteristic, such as "Hello Car" or similar. Upon detection of such an initialization command, the event detection device can, for example, activate a voice assistance system present in the motor vehicle to capture further commands from a user and execute the associated functions. This advantageously also allows a user of the motor vehicle to use the voice assistance function even when the motor vehicle is inactive, in which such a voice assistance function is generally not available.For example, even when the vehicle is locked, the user or owner can use a voice command to open the trunk or easily change the wiper blades on a windshield wiper. For example, a user near the vehicle can give the voice command "wiper blades up" as an example of such a magic command. The vehicle will then first check for the presence of a keyless key belonging to and assigned to the vehicle in the vicinity. If present, the vehicle will automatically move the wiper blades to the maintenance position. A similar procedure can be used with the trunk or a fuel cap or the general operation of any vehicle component using a suitable assigned voice command.The only difference is that a different voice command can be provided for opening the trunk and / or the fuel cap, or for operating another component. In all of these scenarios, it is not necessary for the owner or user of the vehicle to unlock or activate the vehicle beforehand. This means that, for example, voice inputs can also be used outside the vehicle to trigger certain functions.

[0045] Different functions of the motor vehicle can therefore be assigned different keywords and provided, for example, in the form of respective reference characteristics of the event detection device. These can be detected as a specific event by means of the event detection device as described above, in particular using the basic processing unit, the first, second, and third processing modules. When such a keyword is ultimately detected by the third processing module, the microcontroller can be awakened as the fifth processing module, which then triggers a suitable function, e.g., issues a corresponding control command to trigger the vehicle function associated with the keyword.

[0046] The second and third, and optionally fourth, processing modules can each use customized and application-optimized hardware accelerators for neural networks to meet the requirements for low power consumption while maintaining robust recognition. Furthermore, the microcontroller can perform other tasks. The architecture of the neural networker and its associated parameters can be updated, for example, via the microcontroller. The microcontroller can be used for control tasks and can be freely programmed to increase the functionality and flexibility of the system, i.e., the event detection device. The neural networks used in the second and / or third and / or fourth processing modules can be used directly to recognize human speech and words, without any further post-processing.Alternatively, Hidden Markov Models can be used together with the Viterbi algorithm to recognize individual words.

[0047] In addition to the use of microphone signals, the invention can also be used with other sensors in the vehicle that represent or are intended to represent an "always-on" application. These could be, for example, monitoring sensors for batteries or hydrogen tanks, or other applications explained in more detail later.

[0048] In general, a further advantageous embodiment of the invention provides that for a plurality of different, predetermined events to be detected by the event detection device, at least one associated reference characteristic is specified, wherein the event detection device is designed to determine, by comparing the derived sound characteristic of the detected sound signal with the plurality of reference characteristics, from which of the plurality of events the sound detected as the sound signal originates. Based on the detected sound signal and its analysis, which can be carried out as described above, not only can a specific event be detected, but it can also be classified as to which of several possible, predetermined events it is.Furthermore, for each event to be detected by the event detection device, not only one assigned reference characteristic can be specified, but possibly also several different reference characteristics, as already described above. The event detection device can therefore, for example, gradually compare the sound characteristic derived from the detected sound signal with the multiple reference characteristics, which can be implemented, for example, within the framework of the second and / or third processing module described above, and thus easily determine which of the multiple possible events is involved. In particular, this event can be determined by the correspondence between the reference characteristic assigned to this event and the derived sound characteristic being maximum and / or exceeding a predetermined threshold.However, more complex criteria may also be provided.

[0049] This advantageously makes it possible to detect not just one event, but many different events, and even differentiate between them. This also advantageously makes it possible, for example, to provide a dedicated function for each event that is triggered upon detection of the respective event. In other words, the event detection device can trigger different reactions to different events.

[0050] It is a further very advantageous embodiment of the invention if the event represents at least one of the following:

[0051] - An environmental event, e.g. a weather event, for example hail and / or storm or a roof avalanche and / or rain and / or storm damage, or even an earthquake, or similar;

[0052] - Vandalism to the motor vehicle, for example breaking or breaking a mirror, paint damage, for example due to scratching, puncturing of one or more tires, wheel theft or jacking up of the motor vehicle, breaking windows or throwing objects at the car, trailer theft,

[0053] - an accident, in particular contact of the motor vehicle with a person and / or an object external to the motor vehicle and / or damage to the motor vehicle, for example a parking bump caused by another vehicle parking in the vehicle, a shopping trolley bump, a trailer rolling against the motor vehicle, a collision with another road user, for example with a cyclist who rides into the parked motor vehicle,

[0054] - a faulty condition of a component internal to the vehicle, for example a gearbox, an electric motor, an internal combustion engine, one or more generators, pumps, etc.,

[0055] - a living being located in the motor vehicle, in particular one that is enclosed, for example a child or a dog still located in the motor vehicle, in particular in a locked motor vehicle, - a voice command, in particular a voice-based initialization of a voice input for the voice-based operation of a vehicle component of the motor vehicle,

[0056] - a specific vote of a specific person to identify that person as an authorized person;

[0057] - Cell defects in a motor vehicle battery, e.g. thermal runaway and outgassing of a battery cell in a high-voltage battery in the motor vehicle;

[0058] - a fire in a neighboring vehicle;

[0059] - an activated siren, e.g. of an ambulance, a fire brigade, an emergency medical vehicle, a police car, etc.;

[0060] - a break-in into the motor vehicle.

[0061] This means that monitoring can be provided in a particularly advantageous way in numerous different situations, with numerous different events being detected and these can be differentiated from one another, i.e. classified. It is particularly advantageous to be able to detect environmental influences such as weather events. Hail or other weather events in particular can cause very significant damage to a vehicle. Early detection of the occurrence of such a weather event is particularly advantageous in order to initiate measures well in advance and thus before (severe) damage to the vehicle occurs, for example by sending a warning message to the owner of the vehicle, e.g. via an app, and / or driving autonomously into a garage or carport, automatically closing open vehicle windows, or similar.Vandalism or other damage to the vehicle, for example, shopping carts being bumped into or parking spaces being bumped by other vehicles, can also be reliably detected. In this case, too, measures can be initiated, for example the user or owner can be informed immediately or environmental sensors, such as the vehicle's exterior cameras, can be activated to record the accident situation, or similar. An alarm system can also be activated to scare away vandals. Likewise, a break-in into the vehicle can be detected using the event detection device. In this case, too, the alarm system can be triggered and / or the user can be informed as a measure. Furthermore, it is advantageously possible to provide component monitoring of internal components and parts.This enables early detection of defects in production machines such as engines, transmissions, generators, pumps, etc. This can occur not only when the vehicle is inactive, but also, and above all, when the vehicle is active. Certain error states or defects in a transmission, the electric motor, or other components can thus also be advantageously detected based on acoustic or sound signals. This allows malfunctions or defects to be identified very early and, if necessary, major damage to these components to be prevented through timely repair. As a reaction measure, a recommendation can be issued to the user of the vehicle to carry out maintenance or to visit a workshop. The event detection device can also advantageously implement occupant detection, particularly while the vehicle is parked and locked.Children or dogs forgotten in the vehicle can also be reliably detected and appropriate action, such as a warning, can be initiated.

[0062] For example, it can also be detected when the motor vehicle is being washed, e.g. by hand. This can also be detected based on the typical sound characteristics. The event detection device can then take measures such as closing any open windows of the vehicle, automatically adjusting the mirrors, or similar. The event detection device can also detect magic commands, i.e. voice commands, in particular key voice commands, as described above. The event detection device can also be used for voice recognition. If a specific voice of a specific person is recognized who represents an authorized person or is defined as such, the automatic unlocking of the vehicle and / or enabling of the actual voice recognition can be triggered or activated as a reaction measure.In addition, the event detection device can also be designed to detect a siren as a specific event. If a siren is detected, the driver can be warned and / or an autonomous vehicle can be stopped and the system can wait until the other vehicle with the siren has moved away. Numerous other possible applications are conceivable, not all of which can be listed here. For example, acoustic detection and identification of cell defects in a battery, e.g. thermal runaway and outgassing of a battery cell in a high-voltage battery of an electric vehicle, detection of a fire in a neighboring vehicle, earthquake detection, for example during a charging process of the motor vehicle to charge a high-voltage battery of the motor vehicle, and so on. Possible measures in this case include aborting the charging process and / or switching off the battery and / or warning the user.

[0063] In principle, the detectable events can be grouped into environmental events and internal events, whereby the environmental events have their origin in the environment of the vehicle, e.g. weather events, vandalism, accidents, earthquakes, etc. The detection of such environmental events is particularly advantageous, as they are often unpredictable and can cause great damage, so that monitoring, especially permanent monitoring, is very advantageous in this case.

[0064] In a further advantageous embodiment of the invention, the event detection device, in particular the aforementioned microcontroller, is designed to trigger at least one function associated with the event depending on the detection of the event, in particular depending on which event was detected. This is particularly advantageous because the ability to distinguish between different events now allows for particularly good situational adaptation. Thus, there is no need to trigger a blanket response to the detection of such an event; instead, an event-specific response can be triggered.

[0065] Examples of such reactions, which have sometimes also been referred to as measures or reaction measures, have already been mentioned above for the corresponding events. In general, such a reaction can, for example, comprise the driver or owner of the motor vehicle being notified, for example by a push message in an app, in particular a motor vehicle app, on a mobile communication device of the driver, owner or user. Furthermore, as a reaction, an alarm system can be activated or triggered to issue an alarm, an e-call, i.e. an automatic emergency call, can be triggered, hazard lights can be triggered, etc. Furthermore, the vehicle can, under certain circumstances, also be moved autonomously in response, for example by being reparked, in particular to avoid an accident, or by driving out of or into a garage.As a further reaction, for example, a charging process can be terminated and / or cameras and / or recording can be activated, a suggestion for a workshop visit or a workshop can be issued in an output device of the motor vehicle, in particular in the MMI (multimedia interface) of the motor vehicle, or even a suggestion for an authorized dealer, for example to repair detected paint damage or for a mirror repair or other detected damage, or a corresponding action for access and / or maintenance on the vehicle can be carried out by voice commands, for example the fuel filler flap can be opened, wiper blades can be moved to a certain position, the trunk can be opened, the exterior mirrors can be folded in, and so on.

[0066] Furthermore, the invention also relates to a motor vehicle with an event detection device according to the invention or one of its embodiments. The advantages mentioned for the event detection device according to the invention and its embodiments apply equally to the motor vehicle according to the invention.

[0067] The motor vehicle according to the invention is preferably designed as a motor vehicle, in particular as a passenger car or truck, or as a passenger bus or motorcycle.

[0068] Furthermore, the invention also relates to a method for detecting at least one specific event relating to a motor vehicle when the motor vehicle is inactive. A sound-generating event is detected as the event, wherein the sound generated by the event is recorded as a sound signal by a sound sensor, a sound characteristic is determined from the recorded sound signal, and the event is detected depending on the determined sound characteristic. Here, too, the advantages described for the event detection device according to the invention and its embodiments apply equally to the method according to the invention.

[0069] The invention also includes further developments of the method according to the invention that have features already described in connection with the further developments of the event detection device according to the invention. For this reason, the corresponding further developments of the method according to the invention are not described again here.

[0070] The invention also encompasses combinations of the features of the described embodiments. The invention therefore also encompasses implementations that each comprise a combination of the features of several of the described embodiments, unless the embodiments are described as mutually exclusive.

[0071] Exemplary embodiments of the invention are described below. Shown are:

[0072] Fig. 1 is a schematic representation of a motor vehicle with an event detection device according to an embodiment of the invention;

[0073] Fig. 2 is a schematic representation of the event detection device and its functional units according to an embodiment of the invention;

[0074] Fig. 3 is a schematic representation of a component arrangement providing the event detection device according to an embodiment of the invention; and

[0075] Fig. 4 is a flowchart illustrating a method for detecting a specific event according to an embodiment of the invention.

[0076] Fig. 1 shows a schematic representation of a motor vehicle 10 with an event detection device 12 according to an exemplary embodiment of the invention. In this example, the event detection device 12 has a sound sensor 14, which in this example is embodied as a microphone 16, and a processing device 18 based on FD-SOI technology. This FD-SOI semiconductor technology is particularly advantageous for achieving "ultra-low power," i.e., extremely power-saving operation of the event detection device 12. This advantageously enables continuous operation of this event detection device 12, i.e., not only in the active state of the motor vehicle 10, but also in an inactive state ZI of the motor vehicle 10, as shown in this example.The low energy consumption achieved by the microphone-based event detection device 12 is not only possible by using the FD-SOI technology, but additionally or alternatively, and above all primarily, by a cascaded wake-up strategy of individual modules of the processing device 18, as will be described in more detail later.

[0077] In other words, this makes it possible to provide a power-loss-optimized microphone-based event detection device 12 in battery-operated continuous operation for the detection of specific events E1, E2, such as environmental events E1 or speech events E2. The event detection device 12 can therefore advantageously be supplied with energy even in the inactive state ZI of the motor vehicle 10 by an energy storage device 20 of the motor vehicle, for example, a low-voltage battery, such as a 12-volt battery of the motor vehicle 10. Nevertheless, due to the extremely energy-saving operation of the event detection device 12, there is no need to fear that the low-voltage battery 20, which is not recharged in the inactive state ZI of the motor vehicle 10, will be excessively discharged. Furthermore, other energy sources, e.g., solar cells, energy harvesting methods with supercapacitors, etc., can also be used.as a power supply for the sensor, ie, for the event detection device 12. Thus, the event detection device 12 can advantageously provide energy-saving monitoring of the motor vehicle 10 around the clock.

[0078] Another very great advantage of the event detection device 12 is that it not only allows specific events E1, E2 to be detected, but also allows the type of event E1, E2 to be determined. Two different events E1, E2 are shown as examples in Fig. 1. Firstly, as an example of a first event E1, there is an environmental event E1, for example hail E1, which strikes the motor vehicle 10. The impact of the hailstones on the motor vehicle 10 causes a specific first sound signal S1, in this example in the form of sound transmitted through the air. This sound signal S1 can be recorded accordingly by the microphone 16 and forwarded to the processing device 18. This can derive a specific first sound characteristic C1 from the recorded sound signal S1, e.g.generate a frequency spectrum as an example of such a sound characteristic C1 by means of Fourier transformation. This generated frequency spectrum C1 can then be compared with reference characteristics R1, R2 stored in a memory 22 of the processing device 18. In this example, a first reference characteristic R1 is assigned to the first event E1, and a second reference characteristic R2 is assigned to a second event E2. These can also be frequency spectra that are characteristic of the respective events E1, E2. Several assigned reference characteristics R1, R2 can also be stored for each event E1, E2 to be detected.Furthermore, these do not have to be fixed spectra; rather, the processing device 18 can optionally also implement a learning method using an artificial neural network, so that the processing device 18 is designed, for example, in a learning phase, to automatically extract or derive such reference characteristics R1, R2 from detected sound signals S1 and to store and / or modify them as such references and to use them for the future detection of various events E1, E2. Variations occurring in newly detected events E1, E2 concerning such characteristics can then also be stored as a modification or adaptation of the reference characteristics R1, R2.

[0079] If there is a match when comparing the derived characteristic C1 with one of the stored reference characteristics R1, R2 or if the similarity to one of the stored spectra R1, R2 is sufficiently great, which can be determined, for example, by a threshold value, the corresponding event assigned to the reference characteristic R1, R2, in this example the environmental event E1, can be considered detected. As a further example, a second event E2 in the form of a speech event is shown here. The utterance of speech E2 by a person P also generates a corresponding, second sound signal S2, in particular in this example again a sound signal transmitted through air, which is recorded by the microphone 16.The processing device 18 in turn generates a derived second sound characteristic C2 from this detected sound signal S2, in particular again in the form of a frequency spectrum, and compares this frequency spectrum C2 with the stored reference spectra R1, R2. The evaluation of the sound signal S2 can vary from application to application. The event detection device 12 can, for example, implement voice recognition in order to identify the person P, and / or keyword recognition and / or speech recognition in order to activate a vehicle function of the motor vehicle 10 depending on the recognized keyword. A corresponding reference characteristic R2 can be provided for each keyword to be recognized and / or as a voice characteristic of a person P to be identified.

[0080] If there is sufficient agreement with one of the stored reference spectra R1, R2, the event assigned to the reference spectrum is deemed to have been detected, as in this example is the second event E2, namely the speech event E2, e.g. the speaking of a certain keyword. In this way, the processing device 18 can reliably detect various events E1, E2 and even differentiate and classify them according to their type. This also makes it possible to trigger corresponding reactions, in particular even event-specific reactions. For example, each event E1, E2 that is detectable by the event detection device 12, i.e. for which, for example, at least one reference characteristic R1, R2 is stored, can be assigned a reaction or function to be triggered, which is then triggered accordingly upon detection of the event E1 or E2.

[0081] To further increase the energy efficiency of the event detection device 12, it is also provided that it can be operated in different modes. In particular, it can be in a sleep mode in which only a pre-filtering and / or sound pressure detection of received sound signals takes place by the processing device 18. This pre-filtering or pre-analysis can be kept relatively elementary and simple, so that it does not require much energy. Only when a potential candidate for a possible event has been found on the basis of this pre-analysis can the processing device 18 or a main analysis unit of the processing device 18 be woken up orThe processing device 18 can switch to its active analysis mode M2, in which, for example, the corresponding spectra C1, C2 are generated from the detected sound signals S1, S2, and a more in-depth comparison with the stored reference spectra R1, R2 is performed. This will now be explained in more detail with reference to Fig. 2.

[0082] Fig. 2 shows a schematic representation of the event detection device 12 and its individual functional components. The event detection device 12 can be divided into several functional blocks BO, B1, B2, B3, B4, B5. These blocks BO, B1, B2, B3, B4, B5 or the functions or processing tasks AO, A1, A2, A3, A4 executed or executable in the respective blocks BO, B1, B2, B3, B4, B5 are executed by the processing device 18. By way of example, the microphone 16 itself is also part of a block BO. The processing device 18 can be divided into several units or modules, each of which is assigned such a block BO, B1, B2, B3, B4, B5. The processing device 18 comprises a basic processing unit 18a, which, together with the microphone 16, in particular a MEMS microphone 16, is assigned to a first block BO.The base processing unit 18a performs mono- or multi-band sound pressure detection AO with very low energy consumption when switched on. For this purpose, the microphone 16 transmits detected sound signals in the form of a data signal DSO to the base processing unit 18a. The first block B0, in particular the microphone 16 and the base processing unit 18a, are always active in the first idle mode M1 (see Fig. 4). The base processing unit 18a decides, based on the provided data signal DSO, whether a detected sound signal is a potentially relevant signal or not. In the simplest case, this can be decided based on the sound pressure level of this signal. If, for example, a predetermined threshold value for the sound pressure level is exceeded, the base processing unit 18a wakes up a system controller 26, which is assigned to the second block B1, by means of a control signal, in particular a wake-up signal WS.As a result, the event detection device 12 switches to active mode M2 ​​(see Fig. 4). In general, in Fig. 1, data signals DS0, DS1, DS2, DS3, DS4, which carry data, particularly sound data, containing information about the detected sound signal, are represented by solid arrows according to their transmission direction, and control signals, particularly in the form of wake-up signals WS, are represented by dashed arrows, also in the possible transmission directions.

[0083] The processing device 18 also includes several processing modules 18b, 18c, 18d, 18e, 18f, which can only be activated in the active mode M2 ​​of the event detection device 12. Conversely, however, not all of these processing modules necessarily have to be active in the active mode M2 ​​of the event detection device 12. Rather, these processing modules can implement various processing levels, which are only activated and require power dissipation once a corresponding detection has occurred in the previous level.

[0084] If, for example, the system controller 26 has been woken up by the basic processing unit 18a, it wakes up a first processing module 18b via a corresponding WS for digital and / or analog feature extraction from the detected sound signal, which is provided to the first processing module 18b by the microphone 16 via the data signal DS1, in particular in the form of an audio signal. The preprocessing A1 of the audio signal by the first processing module 18b can take place in the form of digital preprocessing A1. The digital preprocessing A1 in block B1 can be replaced by analog preprocessing A1. Suitable for optional preprocessing A1 are, among other things, a bandpass filter, a Fourier transformation (also using the Goertzel algorithm), a logarithmic Fourier transformation, and Mel Frequency Cepstral Coefficients.Reduced calculation accuracy can be achieved through appropriate implementation if the subsequent classification algorithms are sufficiently robust. In addition, the sampling rate can be dynamically adjusted depending on the application.

[0085] The features extracted during preprocessing A1, which can represent the aforementioned sound characteristics C1, C2, can then be provided to further processing modules 18c, 18d, 18e, provided they are awakened. After preprocessing A1, the system controller 26 first awakens a second processing module 18, which is assigned to the third block B2. Based on the features provided by the first processing module 18b, transmitted via the data signal DS2, the second processing module carries out speech recognition as processing task A2. If no speech is detected, the event detection device 12 returns to sleep mode M1. If, on the other hand, the presence of speech in the sound signal is detected, or that the sound signal is a speech signal or speech signal, i.e.about spoken language, the system controller 26 wakes up the next processing module 18d, in this example a processing module for keyword detection, which is assigned to the fourth block B3. The third processing module 18d is therefore designed to detect certain, predetermined keywords as processing task A3. For this purpose, the third processing module 18d also uses the features extracted by the first processing module 18b, transmitted via the corresponding data signal DS3. If no keyword is recognized, the event detection device 12 switches back to sleep mode M1. If a keyword is recognized, the system controller 26 optionally wakes up a fourth processing module 18e, which is assigned to the fifth block B4 and which can perform automatic speech recognition as processing task A4.For this purpose, the fourth processing module 18d also uses the features extracted by the first processing module 18b, transmitted via the corresponding data signal DS4. For some applications, however, the recognition of keywords may already be sufficient to trigger a corresponding assigned function F1 (see Fig. 4), for example, depending on which keyword was recognized. If the fourth processing module 18e does not detect any relevant, significant content, the event detection device 12 can return to sleep mode M1 without triggering any function.

[0086] A fifth processing module 18f, provided by a microcontroller 28, is assigned to the sixth block B5. Such a microcontroller 28 can also be optionally integrated. This can be used for control tasks A5 and can be freely programmed to increase or expand the range of functions and the flexibility of the system, i.e., the event detection device 12. This can, for example, trigger the above-mentioned function F1. This described cascading of functions AO, A1, A2, A3, A4, A5, with each increasing processing complexity and - associated with this - ever higher power consumption, means that the entire cascade only needs to be woken up in rare cases. Overall, this results in a significant reduction in power consumption.

[0087] Fig. 3 shows a schematic exploded view of a microphone-based event detection device 12 according to an embodiment of the invention. The event detection device 12 comprises a seal 30, a filter assembly 32, which may comprise one or more filters, a housing 34 with a housing base 34a and a cover 34b, a circuit board 36 with components, and an insulating adhesive 38. The components of the processing device 18 can be located on the circuit board. The circuit board 36 is arranged in the housing 34 and fastened thereto via the adhesive 38. The housing 34 is resistant to harsh environmental conditions. The round diaphragm of the microphone 16 is located within the filter assembly 32 and is replaceable, for example, using the twist-lock principle. The microphone housing 34 or the entire event detection device 12 can be mounted, for example, on a bumper in order to record realistic sound data.The optimized housing 34 or the overall arrangement of the components of the event detection device 12 enables a very good signal-to-noise ratio, a high resonance frequency and a high overload sound pressure level.

[0088] Fig. 4 shows a flowchart illustrating a method for detecting specific sound-generating events E1, E2 according to an exemplary embodiment of the invention. The method starts in step S10, in which the event detection device 12 is in a first mode M1, which represents a power-saving sleep mode M1. In this first mode M1, the event detection device 12 detects sound events, in particular continuously, via the sound sensor 14, for example the inertial sensor 16. Not all of these sound events are relevant and can be assigned to an event to be detected. Accordingly, in this first mode M1, the event detection device 12 first checks in step S12 whether the detected sound signals are relevant sound signals or irrelevant sound signals. This pre-filtering can be carried out by simple electrical signal filtering or sound pressure level detection.If these are irrelevant sound signals, the process goes to step S10, so the event detection device 12 remains in the first mode M1 and continues to detect sound signals from the environment 24 (cf. Fig. 1 ) of the motor vehicle 10. Only when a detected sound signal has been detected as possibly relevant in step S12 does the event detection device 12 switch in step S14 to a second mode M2, which represents an active analysis mode of the event detection device 12. In this second mode M2, the power consumption is on average higher than in the first mode M1. However, due to the FD-SOI technology, in which the processing device 18 is optionally designed, the power consumption is still relatively low here too. In addition, this second mode M2 ​​can be subdivided into further submodes with different energy levels by the cascaded activation of the processing modules, as described for Fig. 2.

[0089] In this second mode M2, a sound characteristic, in this example the first sound characteristic C1 described in Fig. 1, is now generated in step S16 on the basis of the possibly relevant received sound signal, for example by means of a Fourier transformation, in particular by the first processing module 18b.

[0090] Subsequently, the generated sound characteristic C1 is compared with the reference characteristics R1, R2 stored in the memory 22 in step S18. Such a comparison can be carried out on several levels using different reference characteristics. For example, a first comparison with first reference characteristics, e.g., frequency spectra that are characteristic of speech, can be carried out by the second processing module. If there is sufficient agreement and thus the detection of speech, a second comparison with further second reference characteristics, e.g., frequency spectra or other frequency properties or properties of the spectrum that are characteristic of certain words, in particular keywords, can be carried out by the third processing module.The comparison step S18 can therefore be subdivided into further sub-steps, carried out by the respective processing modules 18b, 18c, 18d, 18e. Within the scope of some of these sub-steps, a further query S18' can be made as to whether the detection in the respective sub-step was successful, e.g. a sufficient match with one of the reference characteristics. If this is not the case in one of the sub-steps, the cascade is aborted and the process can proceed to step S10, and the event detection device 12 returns to the inactive state. Otherwise, the process proceeds to step S20. If a match with the first reference characteristic R1, in particular a respective first reference characteristic in a respective one of the sub-steps in step S18, is determined, the first event E1, which is characterized by the first reference characteristic R1, is now considered to have been detected in step S20.Subsequently, the event detection device 12 triggers a function F1 in step S22. This function can represent a function F1 specifically assigned to the detected event E1. The method can then begin again from the beginning at step S10.

[0091] Overall, the examples demonstrate how the invention can provide a power-loss-optimized, microphone-based event sensor with a cascaded wake-up strategy for continuous battery-powered operation for event detection. A vehicle can be very effectively monitored electronically by the driver or by sensors during active states such as charging or driving. However, if the driver is not in the vehicle or in the vicinity of the vehicle, the electronics, namely the event detection device, can take over this monitoring. To prevent the 12-volt battery or other energy storage device from being over-discharged in a parked vehicle, these electronics, in particular the event detection device and its processing device, are designed to be very power-efficient.By monitoring the car body for characteristic vibrations or by acoustic monitoring, various signals can be advantageously detected. These signals can include, for example, vibrations caused by accidents, such as parking bumps, or environmental influences, such as hail or storms, or even acoustic signals, such as hail, tire theft, paint damage caused by deliberate scratching, and so on. This allows the user to have the parked car monitored 24 / 7, i.e. continuously, using the event detection device. For this purpose, the detected signals in the time domain, i.e. the measured acoustic signal, can be advantageously measured and characterized in a power-saving manner. This is achieved primarily through pre-filtering for relevance and the cascaded wake-up strategy of individual processing modules.Subsequently, further vehicle functions can be initiated. The acoustic signal can thus be detected using a power-saving event detection device even in vehicle states in which the vehicle or the 12-volt battery is not being charged, without resulting in increased idle current consumption in the vehicle. This enables continuous monitoring, providing maximum safety and convenience for the user, and allows the user to implement many functions that would otherwise be unavailable due to excessive power consumption of the parked vehicle. Overall, a power-saving detection methodology can be provided, new semiconductor technologies can be used to achieve extremely power-efficient operation, and many new user functions can be implemented with little additional effort or cost.The power-saving microphone-based event detection system barely increases the idle power consumption of the parked vehicle. The first filter stage for power-saving event evaluation allows additional vehicle functions to be activated as needed, for example, for specific actions or for further plausibility checks, thus opening up new functions for the user.

[0092] Microphones in continuous operation have long been used in the consumer sector to detect voice commands and trigger appropriate actions. Examples of this are Google Alexa or Amazon Echo. However, for devices that are operated from the mains, the power loss of the device is not a limiting factor for continuous operation. The present invention and / or its embodiments relate to the use of microphones in the context of a car in parked and moving operation or in other contexts that require battery operation. For conventional parking operation, the power loss of a conventional microphone in continuous operation is a significant problem: It cannot be guaranteed that the vehicle battery will not be significantly discharged even after weeks and months of continuous operation. However, this can be achieved by the invention and / or its embodiments.On the other hand, such a microphone offers the possibility of covering a wide range of new applications, which are now also possible in continuous operation. For example, the driver can instruct their vehicle to open the trunk or lower the windows using voice commands. Driver authentication and vehicle unlocking using only voice commands are also conceivable, possibly in conjunction with the existing key and wireless technology. Microphones or microphone arrays with very low power dissipation can be continuously operated by a battery and can perform additional functions in the car or other devices.

[0093] The property of very low power consumption is achieved in particular by the following measures: Only a small part of the electronics and a physical microphone (of a possible microphone array) are always kept active; A constant threshold detection of the volume is carried out; Only when the volume threshold is exceeded are further circuits switched on, which analyze the detected event in more detail; Only when this analysis is finished and a predefined event has been detected, are further actions carried out, e.g. by switching on additional functions.

[0094] LIST OF REFERENCE SYMBOLS:

[0095] 10 motor vehicle

[0096] 12 Event detection device

[0097] 14 Sound sensor

[0098] 16 Microphone

[0099] 18 Processing facility

[0100] 18a Basic processing unit

[0101] 18b first processing module

[0102] 18c second processing module

[0103] 18d third processing module

[0104] 18e fourth processing module

[0105] 18f fifth processing module

[0106] 20 energy storage units

[0107] 22 storage

[0108] 24 Surroundings

[0109] 26 System Controllers

[0110] 28 microcontrollers

[0111] 30 Seal

[0112] 32 filter arrangement

[0113] 34 housings

[0114] 34a Lower housing part

[0115] 34b cover

[0116] 36 circuit boards

[0117] 38 Adhesive

[0118] AO processing task

[0119] A1 processing task

[0120] A2 processing task

[0121] A3 processing task

[0122] A4 processing task

[0123] A5 Processing task

[0124] BO function block B1 function block

[0125] B2 Function block

[0126] B3 Function block

[0127] B4 Function block

[0128] B5 Function block

[0129] C1 first sound characteristic

[0130] C2 second sound characteristic

[0131] DSO data signal

[0132] DS1 data signal

[0133] DS2 data signal

[0134] DS3 data signal

[0135] DS4 data signal

[0136] E1 first event

[0137] E2 second event

[0138] F1 function

[0139] M1 first mode

[0140] M2 second mode

[0141] P Person

[0142] R1 first reference characteristic

[0143] R2 second reference characteristic

[0144] 51 first sound signal

[0145] 52 second sound signal

[0146] WS wake-up signal

[0147] ZI inactive state

[0148] S10 Step

[0149] S12 Step

[0150] S14 Step

[0151] S16 Step

[0152] S18 Step

[0153] S18' Step

[0154] S20 Step

[0155] S22 Step

Claims

CLAIMS: 1 . Event detection device (12) for a motor vehicle (10), wherein the event detection device (12) is designed to detect at least one specific event (E1, E2) in an inactive state (ZI) of the motor vehicle (10), characterized in that the event detection device (12) has a processing device (18) and at least one sound sensor (14, 16) which is designed to detect sound (S1, S2) transmitted via air, wherein the event detection device (12) is designed to - to detect an event (E1, E2) generating sound (S1, S2) as the event (E1, E2), - to detect the sound (S1, S2) generated by the event (E1, E2) as a sound signal (S1, S2) by means of the sound sensor (14, 16), - to determine a sound characteristic (C1, C2) from the detected sound signal (S1, S2) by means of the processing device (18), and - to detect the event (E1, E2) depending on the determined sound characteristic (C1, C2).

2. Event detection device (12) according to claim 1, characterized in that the event detection device (12) has at least one FD-SOI semiconductor component, in particular as part of the processing device (18), which comprises, for example, an integrated circuit with an FD-SOI substrate.

3. Event detection device (12) according to one of the preceding claims, characterized in that the event detection device (12) has a sleep mode (M1) and at least one active mode (M2), wherein the event detection device (12) is designed to generate sound signals (S1, S2) in the sleep mode (M1). to detect and classify as relevant or not relevant according to at least one specific relevance criterion, and in the case of a classification of a sound signal (S1, S2) as relevant, to switch to the active mode (M2), in particular wherein a power consumption by the event detection device (12) in the idle mode (M1) is lower than in the at least one active mode (M2).

4. Event detection device (12) according to one of the preceding claims, characterized in that the processing device (18) has a basic processing unit (18a), in particular a sound pressure detection unit (18a), wherein the processing device (18) is set up such that the basic processing unit (18a) is active in the sleep mode (M1), in particular as the only component of the processing device (18), and classifies the detected sound signal (S1, S2) as relevant or not relevant according to the determined relevance criterion, in particular wherein the basic processing unit (18a) is designed to provide a wake-up signal (WS) depending on a signal level of the detected sound signal (S1, S2).

5. Event detection device (12) according to one of the preceding claims, characterized in that - the processing device (18) comprises a plurality of processing modules (18b, 18c, 18d, 18e, 18f), - wherein each of the processing modules (18b, 18c, 18d, 18e, 18f) is designed to execute an associated processing task (A1, A2, A3, A4, A5), - wherein each of the processing modules (18b, 18c, 18d, 18e, 18f) has a first idle state and a second active processing state, - wherein the event detection device (12) is configured such that the processing modules (18b, 18c, 18d, 18e, 18f) are in the respective first idle state in the idle mode (M1) of the event detection device (12), and - wherein the event detection device (12) is configured such that the assigned processing task (A1, A2, A3, A4, A5) can be executed by a first of the processing modules (18b, 18c, 18d, 18e, 18f) which is in the assigned second active processing state, while at least a second of the processing modules (18b, 18c, 18d, 18e, 18f) is in the assigned first idle state.

6. Event detection device (12) according to claim 5, characterized in that the event detection device (12) is set up in such a way that the plurality of processing modules (18b, 18c, 18d, 18e, 18f) can be woken up in a predetermined sequence and, when woken up, change from the first inactive idle state to the second active processing state, in particular wherein the event detection device (12) is set up in such a way that a first of the processing modules (18b, 18c, 18d, 18e, 18f), which has executed the associated processing task (A1, A2, A3, A4, A5) in the active processing state, provides a wake-up signal (WS) depending on a result of the executed processing task (A1, A2, A3, A4, A5), depending on which wake-up signal a next one of the processing modules (18b, 18c, 18d, 18e, 18f ) is awakened in order to carry out the processing task assigned to it (A1 , A2, A3, A4, A5).

7. Event detection device (12) according to one of claims 5 or 6, characterized in that the plurality of processing modules (18b, 18c, 18d, 18e, 18f) comprise a first processing module (18b) which, in the active mode (M2) of the event detection device (12) and in its second processing state, is designed to carry out a pre-processing of the detected and relevant classified sound signal (S1, S2) and to forward a result of the pre-processing, in particular the sound characteristic (C1, C2) determined thereby, to to provide at least one second processing module (18c, 18d, 18e), in particular to several of the processing modules (18c, 18d, 18e), if they are awakened to carry out their respective processing task (A1, A2, A3, A4, A5), wherein the preprocessing preferably comprises frequency filtering and / or generating a frequency spectrum (C1, C2) of the sound signal (S1, S2) classified as relevant.

8. Event detection device (12) according to one of claims 5 to 7, characterized in that the plurality of processing modules (18b, 18c, 18d, 18e, 18f) - comprise a second processing module (18c) which is designed to carry out a speech pause detection (A2), in particular depending on the result of the preprocessing (A1), and / or - comprise a third processing module (18d) which is designed to carry out a word recognition (A3), in particular keyword recognition, in particular depending on the result of the preprocessing (A1), and / or; - comprise a fourth processing module (18e) designed to carry out speech recognition (A4), in particular depending on the result of the preprocessing (A1), and / or - a fifth processing module (18f) comprising a microcontroller (28) designed to execute at least one function (A5, F1), in particular depending on the result of the preprocessing (A1) and / or a result of the processing task (A2, A3, A4) executed by the second and / or third and / or fourth processing module (18c, 18d, 18e).

9. Event detection device (12) according to one of the preceding claims, characterized in that the event (E1, E2) represents at least one of the following: an environmental event, e.g. a weather event (E) or an earthquake; vandalism of the motor vehicle (10); an accident; a faulty state of a component internal to the motor vehicle; a living being located, in particular enclosed, in the motor vehicle (10); a voice command (E2), in particular a voice-based initialization of a voice input for voice-based operation of a vehicle component of the motor vehicle (10); a voice (E2) of a specific person (P) for identifying this person as an authorized person; Cell defects of a battery of the motor vehicle (10), e.g. thermal runaway and outgassing of a battery cell of a high-voltage battery of the motor vehicle; a fire of a neighboring vehicle; an activated siren; a break-in into the motor vehicle (10).

10. Method for detecting at least one specific event (E1, E2) relating to a motor vehicle (10) in an inactive state (ZI) of the motor vehicle (10), characterized in that an event (E1, E2) generating sound (S1, S2) is detected as the event (E1, E2), the sound (S1, S2) generated by the event (E1, E2) being detected as a sound signal (S1, S2) by means of a sound sensor (14, 16), a sound characteristic (C1, C2) is determined from the detected sound signal (S1, S2), and the event (E1, E2) is detected as a function of the determined sound characteristic (C1, C2).

Citation Information

Patent Citations

  • Automobile passenger space surveillance method for anti-theft alarm

    DE19642607A1

  • Acoustic alarm system for motor vehicles using echo signals - generated by applying signal to internal passenger compartment with each subjected to spectral analysis and comparison with reference

    DE4226543A1

  • Phase calibration using a neutralised amplifier with varactors

    WO2023062291A1