Ultrasonic interface unit for driver assistance systems

Abstract

The present invention relates to an ultrasonic interface unit for driver assistance systems, which can be communicatively connected to several ultrasonic sensor elements of an ultrasonic sensor array and to a control unit and is configured to convert several ultrasonic sensor data from the ultrasonic sensor elements to a compressed transmission signal by means of a compression transformation and to forward the compressed transmission signal to the control unit.

Description

State of the art

[0001] The present invention relates to an ultrasonic interface unit and an associated ultrasonic sensor system for driver assistance systems, wherein the ultrasonic interface unit is configured to convert multiple ultrasonic sensor data from ultrasonic sensor elements into a compressed transmission signal by means of a compression transformation and to forward the compressed transmission signal to a control unit. A further aspect of the invention relates to a method for processing ultrasonic sensor data with the ultrasonic sensor system.

[0002] Ultrasound-based parking sensors can estimate the distance to a reflecting object by echo detection of the emitted ultrasonic signals. For this purpose, prior art uses filter operations on the receiving end to recognize the signal characteristics from the emitted sound sensor, so that when matches are found, the information is forwarded to the control unit as selected transit times and signal amplitudes. Since only the compressed information of an ultrasonic signal in the form of transit times and signal amplitudes can be processed at the control unit, the processing options are very limited, especially with regard to sensor data fusion. Furthermore, it is not advisable, or in some cases technically impossible, to transmit the complete raw signal, as the network traffic at the correspondingly high sampling frequencies is considerable.In particular, the proliferation of ultrasonic sensor arrays, which comprise a large number of ultrasonic sensor elements, complicates the transmission of ultrasonic sensor data from the individual elements within the array. It would be desirable to expand the processing capabilities for ultrasonic sensor data, thereby broadening the application possibilities of data-driven approaches and enabling better environmental perception by the ultrasonic sensor system. Disclosure of the invention

[0003] The ultrasound interface unit according to the invention, comprising the features of claim 1, the associated ultrasound sensor system comprising the features of claim 7, and the method for processing ultrasound sensor data comprising the features of claim 9, offer the advantage that the processing capabilities of ultrasound sensor data can be improved, enabling enhanced use of the data in driver assistance systems. This is achieved according to the invention with an ultrasound interface unit for driver assistance systems, which can be communicatively connected to several ultrasound sensor elements of an ultrasound sensor array and to a control unit. The ultrasound interface unit is configured to convert multiple ultrasound sensor data from the ultrasound sensor elements into a compressed transmission signal by means of a compression transformation and to forward the compressed transmission signal to the control unit.By combining and compressing multiple ultrasonic sensor data using compression transformation, a high data transmission rate can be achieved from the ultrasonic sensor elements of an ultrasonic sensor array to a control unit. Similarities in the ultrasonic sensor data of neighboring ultrasonic sensor elements can be used to increase the compression factor. The ultrasonic sensor data are preferably the raw signals from the ultrasonic sensor elements, which can be in the form of analog or digital signals. An ultrasonic sensor array describes a system of multiple ultrasonic sensor elements on a small area, which, for example, enables the use of beam steering, and should not be confused with the arrangement of several physically separate ultrasonic sensors on a bumper. Compression transformation can compress the ultrasonic sensor data losslessly or with loss.When compressing ultrasonic sensor data with loss, the information loss should preferably be kept as low as possible. The ultrasonic sensor elements are preferably transmit / receive units configured to transmit and receive ultrasonic signals. The ultrasonic sensor data preferably includes a detected echo signal from the ultrasonic sensor elements. Alternatively, the method can also be used for frequencies outside the ultrasonic frequency range.

[0004] The dependent claims describe preferred embodiments of the invention.

[0005] Preferably, the ultrasonic interface unit is arranged on a printed circuit board with the ultrasonic sensor elements of an ultrasonic sensor array. This enables rapid compression of the ultrasonic sensor data into a compressed transmission signal while maintaining a compact design. In particular, the ultrasonic sensor elements are piezoelectric, micromechanical ultrasonic transducers.

[0006] Preferably, the compression transformation includes mathematical feature extraction in the time domain and / or frequency domain. This allows the ultrasonic sensor data to be converted into a two-dimensional feature vector sequence. This sequence can then be integrated into the compression transformation, for example, using a short-time Fourier transform or wavelet transform. Furthermore, the mathematical features can be extracted by filtering, such as using high-pass, low-pass, or band-pass filters. Based on this mathematical feature extraction, irrelevant and time-limited time segments can be removed from the ultrasonic sensor data.

[0007] The compression transformation preferably comprises a local machine learning model, in particular an autoencoder. This can enable efficient compression of the ultrasonic sensor data into a compressed transmission signal.

[0008] The weighting of the local machine learning model is preferably adjustable by the control unit, particularly using a federated learning model. This allows the machine learning model for compression transformation to be continuously optimized. A federated learning model can enable intelligent optimization of the local machine learning models when multiple ultrasonic sensor arrays are distributed around a vehicle and contain different datasets.

[0009] Furthermore, the invention relates to an ultrasonic sensor system comprising an ultrasonic sensor array with multiple ultrasonic sensor elements, a control unit, and a previously described ultrasonic interface unit. The control unit is configured to extract information from a compressed transmission signal of the ultrasonic interface unit. This eliminates the need for the resource-intensive transmission of raw signals to the control unit. The control unit can enable a complete or partial reconstruction of the ultrasonic sensor data. The ultrasonic sensor elements are preferably piezoelectric, micromechanical ultrasonic transducers. More preferably, the ultrasonic sensor system comprises multiple ultrasonic sensor arrays, each of which is assigned an ultrasonic interface unit.

[0010] Preferably, the control unit is configured to extract information from the compressed transmission signal using regression analysis and / or classification analysis. For example, regression analysis can be used to determine the distance to an obstacle and the direction to the obstacle from a received ultrasound signal. Classification analysis can be used, for example, to determine whether a detected obstacle is large, small, traversable, or impassable.

[0011] Furthermore, the invention relates to a method for processing ultrasonic sensor data. In the method, in a first step, several ultrasonic sensor data points are generated from an ultrasonic signal received by several ultrasonic sensor elements of an ultrasonic sensor array. Subsequently, the ultrasonic sensor data are transmitted to an ultrasonic interface unit. On the ultrasonic interface unit, the ultrasonic sensor data are compressed into a compressed transmission signal by means of a compression transformation. The compressed transmission signal is then transmitted to a control unit. Information can be extracted from the compressed transmission signal by means of the control unit.By combining and compressing the ultrasonic sensor data, the resource-intensive transmission of uncompressed data to the control unit can be avoided, and the processing capabilities of the control unit can be expanded using data-driven approaches. Preferably, the compression includes mathematical feature extraction in the time domain and / or frequency domain. This allows the essential features of the ultrasonic sensor data to be extracted in order to compress their data rate.

[0012] Preferably, the compression process incorporates a machine learning model, in particular an autoencoder. The autoencoder can efficiently extract essential features from the ultrasonic sensor data and enable efficient compression. Machine learning models include, in particular, artificial neural networks.

[0013] The weighting of the machine learning model is preferably trained on the control unit and then adjusted on the ultrasound interface unit. This is achieved particularly using a federated learning model. In a federated learning model, the control unit trains a central machine learning model and transfers it to multiple ultrasound interface units. The individual ultrasound interface units then train their local machine learning models using local data. The control unit then combines the local machine learning models and creates a global machine learning model without accessing the local data. This allows for the creation of a robust machine learning model that is locally optimized.

[0014] Preferably, the information is extracted from the compressed transmission signal using regression analysis and / or classification analysis. This allows, for example, a driver assistance system to be reliably provided with high-quality information from the ultrasonic sensor data. Brief description of the drawings

[0015] Exemplary embodiments of the invention are described in detail below with reference to the accompanying drawings. The drawing shows: Fig. 1 a schematic view of an ultrasonic sensor system with an ultrasonic interface unit according to a first embodiment of the invention, Fig. 2 a representation of a method for processing ultrasonic sensor data with the ultrasonic sensor system according to the first embodiment of the invention, Fig. 3 a description of an alternative method for processing ultrasonic sensor data with the ultrasonic sensor system according to the first embodiment of the invention, and Fig. 4 a representation of a further alternative method for processing ultrasonic sensor data with the ultrasonic sensor system according to the first embodiment of the invention. Embodiments of the invention

[0016] Preferably, all identical components, elements and / or units in all figures are provided with the same reference numerals.

[0017] The following refers to the Fig. 1 to 4 an ultrasonic sensor system 1 with an ultrasonic interface unit 2 and a method for processing ultrasonic sensor data 10 on the ultrasonic sensor system 1 are described in detail.

[0018] Fig. Figure 1 shows the ultrasonic sensor system 1, which comprises an ultrasonic sensor array 4 with several ultrasonic sensor elements 3, a control unit 5, and an ultrasonic interface unit 2. The ultrasonic sensor array 4 has four ultrasonic sensor elements 3 arranged in a 2x2 array. This allows ultrasonic signals 7 to be emitted from the ultrasonic sensor array 4 at a defined angle using beam steering, and the angle of incidence of received ultrasonic signals 7 to be determined.

[0019] The ultrasonic sensor elements 3 of the ultrasonic sensor array 4 are arranged on a common circuit board 6 with the ultrasonic interface unit 2. The ultrasonic interface unit 2 is communicatively connected to the ultrasonic sensor elements 3 in order to receive the ultrasonic sensor data 10 from the ultrasonic sensor elements 3.

[0020] The ultrasound interface unit 2 has a compression transformation 11 (in Fig. (2 to 4 shown) which can convert the multiple ultrasonic sensor data 10 into a compressed transmission signal 12. Furthermore, the ultrasonic interface unit 2 is communicatively connected to the control unit 5 in order to forward the compressed transmission signal 12 to it.

[0021] The ultrasonic sensor system 1 is preferably arranged on a vehicle to obtain information for a driver assistance system. Particularly preferably, the ultrasonic sensor system 1 comprises a plurality of ultrasonic sensor arrays 4 with associated ultrasonic interface units 2, which can compress the ultrasonic sensor data of the individual ultrasonic sensor elements 3 and forward the compressed transmission signal 12 to the control unit 5.

[0022] The ultrasonic sensor elements 3 are preferably transmitting / receiving units that can emit and subsequently receive an ultrasonic signal, so that the distance to a reflecting object can be determined from the time of flight to that object.

[0023] Fig. Figure 2 shows a method for processing ultrasonic sensor data 10 generated by several ultrasonic sensor elements 3. The ultrasonic sensor data 10 are transmitted via the ultrasonic interface unit 2 and compressed there into a compressed transmission signal 12 by means of a compression transformation 11. The compression includes, for example, mathematical feature extraction in the time domain and / or the frequency domain. Sections of the ultrasonic sensor data without relevant information can be removed for compression. For example, a time-frequency transformation can be applied as feature extraction. This can improve the robustness of the data processing.

[0024] Compression can also be based on a machine learning model, such as an artificial neural network. This can learn feature extraction for compression. An example of a machine learning model for compressing ultrasonic sensor data is an autoencoder.

[0025] The compressed transmission signal 12 created by the ultrasonic interface unit 2 from the ultrasonic sensor data 10 is then transmitted by the ultrasonic interface unit 2 to the control unit 5.

[0026] The control unit 5 can have a decompression transformation 24 which can fully or partially reconstruct the ultrasonic sensor data 10 from the compressed transmission signal 12. The reconstructed ultrasonic sensor data 10 can then be filtered on the control unit using filters 25 to improve, for example, the signal-to-noise ratio. Matched filters can be used for this purpose. Information 23 can then be extracted from the filtered ultrasonic sensor data 10 on the control unit 5.

[0027] Fig. Figure 3 shows another method for processing the ultrasonic sensor data. Figure 10. This method is similar to the one described in Figure 3. Fig. 2. The procedures shown up to the transmission of the compressed transmission signal 12 to the control unit 5.

[0028] The compressed transmission signal 12 is evaluated on the control unit 5 using a classification analysis 22 in order to extract information 23. The classification analysis 22 has, for example, five classes into which information based on the received ultrasound signal 7 can be classified.

[0029] The classes can, for example, describe different types of obstacles. In the example shown, the received ultrasound signal is assigned to class 2 with a probability of 80 percent. The classification analysis 22 is preferably performed using a machine learning model, in particular by means of a neural network.

[0030] For the extraction of information from the compressed transmission signal 12 using the control unit 5, the following additional features can be provided to the compressed transmission signal 12, for example, the distance to the object, a signal-to-noise ratio of the ultrasonic sensor data 10, a correlation of the ultrasonic sensor data 10, a sound energy of the backscattering and / or spectral features such as centroid, wavelength or spectral fluctuation of the ultrasonic sensor data 10.

[0031] Fig. Figure 4 shows another method for processing the ultrasonic sensor data. The method in Fig. 4 is similar to the procedure in Fig. 2 and Fig. 3 until the compressed transmission signal 12 is transmitted to the control unit 5.

[0032] The compressed transmission signal 12 is in Fig.4 was evaluated using a regression analysis 21 to extract information 23. Here, for example, it was determined from the compressed transmission signal 12 that the received ultrasound signal 7 was reflected by an object at a distance of 1.5 meters, located at a horizontal angle of 15° and a vertical angle of -30° to the ultrasound sensor array 4. The regression analysis 21 is preferably performed using a machine learning model 14.

[0033] Thus, by compressing several ultrasonic sensor data 10 into a compressed transmission signal 12, a large amount of data can be efficiently transmitted to the control unit 5 of the ultrasonic sensor system 1 via the ultrasonic interface unit 2, where it can be efficiently processed, especially using data-driven approaches, to provide information for a driver assistance system, for example.

Claims

[1] Ultrasonic interface unit for driver assistance systems, which is communicatively connectable to several ultrasonic sensor elements (3) of an ultrasonic sensor array (4) and to a control unit (5) and is configured to convert several ultrasonic sensor data (10) of the ultrasonic sensor elements (3) by means of a compression transformation (11) to a compressed transmission signal (12) and to forward the compressed transmission signal (12) to the control unit (5). [2] Ultrasonic interface unit according to claim 1, wherein the ultrasonic interface unit (2) is arranged on a printed circuit board (6) with the ultrasonic sensor elements (3) of an ultrasonic sensor array (4). [3] Ultrasound interface unit according to one of the preceding claims, wherein the compression transformation (12) comprises mathematical feature extraction in the time domain and / or the frequency domain. [4] Ultrasound interface unit according to one of the preceding claims, wherein the compression transformation (11) comprises a local machine learning model, in particular an autoencoder. [5] Ultrasound interface unit according to claim 4, wherein weights of the local machine learning model are adjustable by the control unit (5), in particular by means of a federated learning model. [6] Ultrasonic sensor system comprising an ultrasonic sensor array (4) with multiple ultrasonic sensor elements (3), a control unit (5) and an ultrasonic interface unit (2) according to one of the preceding claims, wherein the control unit (5) is configured to extract information (23) from a compressed transmission signal (12) of the ultrasonic interface unit. [7] Ultrasound sensor system according to claim 6, wherein the control unit (5) is configured to extract information (23) from the compressed transmission signal (12) by means of a regression analysis (21) and / or a classification analysis (22). [8] Method for processing ultrasonic sensor data (10), comprising the steps: - Generating ultrasonic sensor data (10) from an ultrasonic signal (7) received by several ultrasonic sensor elements (3) of an ultrasonic sensor array (4), - Transferring the ultrasonic sensor data (10) to an ultrasonic interface unit (2), - Compressing the ultrasonic sensor data (10) to a compressed transmission signal (12) by means of a compression transformation (11) on the ultrasonic interface unit (2), - Transmitting the compressed transmission signal (12) to a control unit (5), and - Extracting information (23) from the compressed transmission signal (12) using the control unit (5). [9] Method according to claim 8, wherein the compression comprises mathematical feature extraction in the time domain and / or the frequency domain. [10] Method according to one of claims 8 or 9, wherein the compression comprises a machine learning model, in particular an autoencoder. [11] Method according to claim 10, wherein the weighting of the machine learning model is trained on the control unit (5) and subsequently adapted on the ultrasound interface unit (2), in particular by means of a federated learning model. [12] Method according to any one of claims 8 to 11, wherein the information (23) is extracted from the compressed transmission signal (12) by means of a regression analysis (21) and / or a classification analysis (22).

Images