System for identifying noise sources within a vehicle

Abstract

The present invention relates to the field of noise diagnosis in vehicles. The present invention specifically relates to a testing system for identifying root cause of noise in a vehicle. A system 100 for identifying source of noise within a vehicle is provided. The system 100 includes an acquisition module 102 configured to acquire real-time noise of different frequencies within the vehicle and pre-process the noise to generate output pre-processed noise signals related to the acquired noise. Further, the system 100 includes a beamformer 104 configured to generate a plurality of beams based on analog signals converted from the noise by a plurality of components in the vehicle and identify a beam having highest intensity among the plurality of beams to select one beam and output noise signals corresponding to the identified beam. Still further, the system 100 includes a processing module 106 communicatively coupled with the acquisition module 102 and the beamformer 104 and configured to receive and process the pre-processed noise signals from the acquisition module 102, the noise signals corresponding to the identified beam from the beamformer 104 and a plurality of historical noise data 108 from a plurality of vehicles to determine a source of noise associated with the noise from within the vehicle by using a deep learning model to correlate outputs from the acquisition module 102, the beamformer 104 and the plurality of historical noise data 108.

Description

Technical Field

[0001] This invention relates to the field of noise diagnosis in vehicles. Specifically, it relates to a testing system for identifying the root causes of noise in vehicles. Background Technology

[0002] In the automotive industry today, noise diagnostics has recently gained interest due to rising consumer expectations. Previously, engines, transmissions, and tires were researchers' primary concerns, but with the development of automotive noise diagnostics and the advent of electric vehicles, these have become less relevant, and dynamic noise has become more important. Diagnosing the operational status of a vehicle is a crucial part of vehicle maintenance and repair. Therefore, it is necessary to detect vehicle malfunctions or operational problems.

[0003] Recent developments in noise testing have allowed for more objective measurements of vehicle noise. Currently, most vehicle service centers detect noise sources through subjective stethoscope testing. Existing routine techniques, including stethoscope methods or removing each component of the vehicle to inspect it, are labor-intensive and do not provide immediate feedback to service personnel. However, repair specialists have limitations, as they can only limit their testing to powertrain noise testing. Furthermore, it only alerts the driver to problems without predicting possible noise sources or the causes that might be causing the noise at their source. Therefore, the results vary depending on the repair specialist's expertise.

[0004] Prior art EP1703471B1 discloses a system for automatically identifying vehicle operating conditions using a microphone positioned inside the vehicle. The microphone detects sound signals. A database stores speech templates and operating noise templates. A feature extraction module receives the microphone signals and extracts operating noise feature parameters or a set of speech feature parameters from them. A speech and noise recognition module can determine the operating noise template that best matches the extracted operating noise feature parameters and / or the set of speech templates. The speech template best matches the extracted set of speech feature parameters. Attached Figure Description

[0005] Different embodiments of the present invention are disclosed in detail in the specification and illustrated in the accompanying drawings:

[0006] Figure 1 The illustration shows a system for identifying noise sources inside a vehicle according to an embodiment of the present invention;

[0007] Figure 2 This is based on the various aspects of the use of this technology. Figure 1 The system is used to identify examples of noise sources within a vehicle. Detailed Implementation

[0008] Figure 1The diagram illustrates a system 100 for identifying noise sources within a vehicle. System 100 includes an acquisition module 102, a beamformer 104, a processing module 106, and an output module 114. Each component is described in further detail below.

[0009] The acquisition module 102 is configured to acquire real-time noise at different frequencies within the vehicle and preprocess the noise to generate an output preprocessed noise signal associated with the acquired noise. In one embodiment, the acquisition module 102 includes an array of multiple microphones adapted to acquire real-time noise signals at different frequencies within the vehicle.

[0010] In one embodiment, an array of multiple microphones, including at least one directional microphone or multiple directional microphones pointing in different directions, can be used. In this example, the multiple directional microphones improve the reliability of the vehicle noise acquisition procedure and can also provide better localization of such faults if and when an operational malfunction is detected.

[0011] In one embodiment, the acquisition module 102 is further configured to convert the acquired noise signal into an electronic microphone signal. In operation, the noise signal is acquired by the acquisition module 102 within the vehicle and converted into an electronic microphone signal. In one embodiment, the acquisition module 102 may include a processing unit suitable for preprocessing the microphone signal. Specifically, the microphone signal is digitized and quantized by the processing unit. The processing unit may also perform a Fast Fourier Transform (FFT), order analysis, or some other similar transform to convert the digitized microphone signal from the time domain to the frequency domain. However, various other noise signal transformation techniques are conceivable. The processing unit may also apply an appropriate time delay to synchronize the microphone signals received from the plurality of microphones.

[0012] Beamformer 104 is configured to generate multiple beams based on analog signals converted from noise by multiple components in the vehicle. Beamformer 104 is further configured to identify the beam with the highest intensity among the multiple beams to select a beam, and output a noise signal corresponding to the identified beam. In one embodiment, acquisition module 102 is configured to receive raw noise data from the multiple microphones and drive noise-based behavior using beamforming and echo cancellation.

[0013] Furthermore, beamformer 104 is configured to emphasize noise originating from a specific direction, such as noise from the engine compartment, drivetrain, transmission, driver or passenger, or from some other source. In another embodiment, the beamformer is configured to improve the quality of the noise signal in order to improve the reliability of identifying vehicle operating noise.

[0014] In one embodiment, beamformer 104 is configured to receive a pre-processed microphone noise signal from acquisition module 102. In this embodiment, beamformer 104 is further configured to convert the received pre-processed microphone noise signal to obtain a noise intensity value. The noise intensity value is then further provided to processing module 106. Herein, the noise intensity value is used to locate noise sources from different directions within the vehicle.

[0015] Processing module 106 is communicatively coupled to acquisition module 102 and beamformer 104. Processing module 106 is configured to receive preprocessed noise signals from acquisition module 102, noise signals corresponding to identified beams from beamformer 104, and multiple historical noise data 108 from multiple vehicles. Processing module 106 is configured to receive multiple historical noise data 108, which are stored in a memory or accessible from other locations, such as offline storage and cloud storage, or from multiple vehicles.

[0016] The processing module 106 is further configured to process the preprocessed noise signal from the acquisition module 102, the noise signal corresponding to the identified beam from the beamformer 104, and multiple historical noise data 108 from multiple vehicles to determine the noise source associated with the noise from inside the vehicle. The noise source is associated with the noise from inside the vehicle by using a deep learning model to correlate the outputs from the acquisition module 102, the beamformer (104), and the multiple historical noise data 108.

[0017] In one embodiment, processing module 106 is configured to process outputs from acquisition module 102, beamformer 104, and multiple historical noise data sets 108. Here, noise sources associated with noise from within the vehicle are determined by using a deep learning model to correlate the outputs from acquisition module 102, beamformer 104, and multiple historical noise data sets 108. The noise sources associated with noise from within the vehicle can be displayed to the user via output module 114.

[0018] In one embodiment, processing module 106 includes analysis module 110 and prediction module 112. Analysis module 110 is configured to store noise data from acquisition module 102, beamformer 104, and multiple historical noise data 108 from multiple vehicles. In one example, the noise data may include post-processed noise signals, design and operational data of multiple vehicle subsystems, and corresponding tagged noise source components. Analysis module 110 is further configured to analyze the noise data to identify two or more target noise sources in the vehicles. Further, analysis module 110 may include a memory for storing the noise data.

[0019] The prediction module 112 is configured to use a deep learning model to predict the primary noise source in the vehicle based on two or more target noise sources identified in the vehicle. In one embodiment, noise data from multiple vehicles is used to train a deep learning model, such as a CNN, RNN, DNN, or any other machine learning model. In one embodiment, the deep learning model can use different variants of artificial neural networks, such as deep multilayer perceptrons (MLPs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTMs), sequence-to-sequence models, and shallow neural networks including word2vec for word embeddings.

[0020] Figure 2 Based on the various uses of this technology Figure 1 Example processing 200 of the system for identifying noise sources inside a vehicle.

[0021] At step 201, real-time noise at different frequencies within the vehicle is acquired. The acquired noise is further preprocessed to generate an output electrical signal associated with the acquired noise. The noise signal is acquired by acquisition module 102 and converted into an electrical microphone signal, said acquisition module 102 comprising an array of multiple microphones within the vehicle. In one embodiment, acquisition module 102 may include a processing unit suitable for preprocessing the microphone signal. Specifically, the microphone signal is digitized and quantized by the processing unit. The processing unit may also perform Fast Fourier Transform (FFT), order analysis, or some other similar transform to convert the digitized microphone signal from the time domain to the frequency domain.

[0022] At step 202, multiple beams are generated based on analog signals converted from noise by multiple components in the vehicle. The beam with the highest intensity among the multiple beams is identified to select a beam, and a noise signal corresponding to the identified beam is output.

[0023] At step 203, multiple historical noise data 108 from multiple vehicles are received. The historical noise data 108 can be received from a memory or accessed from other locations, such as access from offline storage and cloud storage, or access from multiple vehicles.

[0024] At step 204, noise sources associated with noise from within the vehicle are determined by associating the output electrical signal from the preprocessed noise signal, the noise output signal corresponding to the identified beam, and multiple historical noise data (108) using deep learning. In this embodiment, a deep learning model is employed to learn the noise signals to determine the accurate noise sources associated with noise from within the vehicle. The deep learning model is used to learn embedded features extracted from noise signals from several components of the vehicle. Examples of such models include, but are not limited to, CNNs, RNNs, DNNs, or any other machine learning models. Further, the received multiple noise signals are transformed into spectrograms or hotspots based on deep data analysis and machine learning to train the deep learning model. At step 205, the determined noise sources associated with noise from within the vehicle are displayed.

[0025] The noise signal is received and analyzed to extract two or more embedding features. The extracted embedding features may include noise / sound features such as log-Mel spectrograms (LM), Mel-frequency cepstral coefficients (MFCC), or the raw waveform of the frequency, to train a deep neural network for noise source identification in a vehicle. In one embodiment, the extracted embedding features are used to train a deep learning model, such as a CNN, RNN, DNN, or any other machine learning model.

[0026] It should be understood that the embodiments described above are merely illustrative and do not limit the scope of the invention. Many such embodiments, as well as other modifications and variations to the embodiments explained in the specification, are contemplated. The scope of the invention is limited only by the scope of the claims.

Claims

1. A system (100) for identifying noise sources inside a vehicle, comprising: The acquisition module (102) is configured to acquire real-time noise of different frequencies inside the vehicle and preprocess the noise to generate an output preprocessed noise signal related to the acquired noise. The beamformer (104) is configured to generate multiple beams based on analog signals converted from noise by multiple components in the vehicle, identify the beam with the highest intensity among the multiple beams to select a beam, and output a noise signal corresponding to the identified beam. The processing module (106), communicatively coupled to the acquisition module (102) and the beamformer (104), is configured to receive a preprocessed noise signal from the acquisition module (102), a noise signal corresponding to an identified beam from the beamformer (104), and multiple historical noise data (108) from multiple vehicles, wherein The processing module (106) is further configured to process a preprocessed noise signal from the acquisition module (102), a noise signal corresponding to an identified beam from the beamformer (104), and multiple historical noise data (108) from multiple vehicles, in order to determine noise sources associated with noise from within the vehicles by associating the outputs from the acquisition module (102), the beamformer (104), and the multiple historical noise data (108) using a deep learning model, wherein the processing module (106) includes: The analysis module (110) is configured to store noise signals from the acquisition module (102), the beamformer (104) and multiple historical noise data (108) from multiple vehicles, and to analyze the noise signals from the acquisition module (102), the beamformer (104) and the multiple historical noise data (108) from multiple vehicles to identify two or more target noise sources in the vehicles. as well as The prediction module (112) is configured to use a deep learning model to predict the main noise source in the vehicle based on two or more target noise sources identified in the vehicle.

2. The system (100) according to claim 1, wherein the acquisition module (102) includes an array of multiple microphones, the array of multiple microphones being adapted to acquire real-time noise signals of different frequencies inside the vehicle.

3. The system (100) according to claim 2, wherein the acquisition module (102) is further configured to convert the acquired noise signal into an electric microphone signal.

4. The system (100) according to claim 1, wherein the processing module (106) is configured to receive a plurality of historical noise data (108) stored in a memory or accessed from other locations, wherein access from other locations includes access from offline storage and cloud storage, and access from multiple vehicles.

5. The system (100) according to claim 3, wherein the historical noise data (108) includes post-processed noise signals, design and operation data of multiple vehicle subsystems, and corresponding tagged noise source components.

6. The system (100) according to claim 3, wherein the beamformer (104) is configured to receive a pre-processed microphone noise signal from the acquisition module (102) and convert the received pre-processed microphone noise signal to obtain a noise intensity value.

7. A method (200) for identifying noise sources inside a vehicle, the method comprising: (201) Collect real-time noise at different frequencies inside the vehicle and preprocess the noise to generate an output preprocessed noise signal related to the collected noise; (202) Multiple beams are generated based on analog signals converted from noise by multiple components in the vehicle, and the beam with the highest intensity among the multiple beams is identified to select a beam, and a noise signal corresponding to the identified beam is output. Receive (203) preprocessed noise signal from acquisition module (102), noise signal corresponding to the identified beam from beamformer (104), and multiple historical noise data (108) from multiple vehicles. The preprocessed noise signal from the acquisition module (102), the noise signal corresponding to the identified beam from the beamformer (104), and multiple historical noise data from multiple vehicles (108) are processed. Using deep learning, the outputs from the acquisition module (102), beamformer (104), and multiple historical noise data (108) are correlated to determine the noise sources associated with the noise from inside the vehicle; The noise signal from the acquisition module (102) and the beamformer (104) and the historical noise data from multiple vehicles (108) are stored. Analyze the noise signals from the acquisition module (102), the beamformer (104) and multiple historical noise data (108) from multiple vehicles to identify two or more target noise sources in the vehicles; as well as A deep learning model is used to predict the primary noise source in the vehicle based on two or more target noise sources identified in the vehicle.

Images