Noise reduction processing method and device, vehicle and storage medium

Abstract

The invention discloses a noise reduction processing method and device, a vehicle and a storage medium, and relates to the technical field of intelligent cabins, and the method comprises the steps: carrying out the recognition based on an error signal and an accelerometer signal of the vehicle, and obtaining a noise scene where the vehicle is located; decomposing the accelerometer signal into a plurality of sub-band signals which are not overlapped with one another; for each sub-band signal, determining a noise reduction processing parameter corresponding to each sub-band signal based on the energy analysis and noise scene of the sub-band signal; based on the noise reduction processing parameter corresponding to each sub-band signal, performing adaptive filtering on each sub-band signal in combination with the error signal to obtain a noise reduction signal corresponding to each sub-band signal; and carrying out fusion reconstruction on the noise reduction signals corresponding to the plurality of sub-band signals to obtain a full-band noise reduction signal. According to the invention, adaptive filtering and noise reduction are carried out on each sub-band signal based on the noise scene where the sub-band signal is located, efficient division and treatment and response can be carried out on complex and multi-spectral structure noise, the noise reduction effect is improved, and the vehicle use experience feeling of a user is further improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of intelligent cockpit, and particularly relates to a noise reduction processing method and device, a vehicle and a storage medium. BACKGROUND

[0002] With the rapid development of new energy vehicles, the power system of the new energy vehicles is changed from a traditional internal combustion engine to an electric drive system, which significantly reduces engine noise, so that the originally hidden wind noise, road noise and electronic device operation sound become the main factors affecting the acoustic comfort of the cockpit. The importance of in-vehicle noise reduction technology is increasingly prominent, which not only relates to the driving experience, but also becomes one of the key indicators for measuring the quality of an intelligent cockpit.

[0003] In the related art, the vehicle-mounted active noise reduction can use a microphone to collect a noise signal, generate a reverse sound wave in real time, and cancel noise in a specific frequency band. The vehicle-mounted active noise reduction is often used to suppress wideband noise such as motor howling, road excitation and wind noise. Most of the vehicle-mounted active noise reduction uses full-bandwidth filtering or a small number of fixed frequency division structures to uniformly and adaptively process wideband noise. However, a single full-bandwidth adaptive filter cannot be targeted at complex multi-source and dynamically changing noise spectrum structures (such as road noise and tire noise) in the automobile environment, and it is difficult to obtain an ideal noise reduction effect in overlapping or weak frequency bands, resulting in an unsatisfactory noise reduction effect.

[0004] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] To overcome the problems in the related art, the present disclosure provides a noise reduction processing method, device, vehicle and storage medium.

[0006] According to a first aspect of an embodiment of the present disclosure, a noise reduction processing method is provided, comprising: identifying a noise scene in which a vehicle is located based on an error signal of the vehicle and an accelerometer signal of the vehicle; decomposing the accelerometer signal into a plurality of mutually non-overlapping sub-band signals; for each of the sub-band signals, determining a noise reduction processing parameter corresponding to each of the sub-band signals based on energy analysis of the sub-band signal and the noise scene; performing adaptive filtering on each of the sub-band signals based on the noise reduction processing parameter corresponding to each of the sub-band signals and the error signal to obtain a noise reduction signal corresponding to each of the sub-band signals; performing fusion reconstruction on the noise reduction signals corresponding to the plurality of sub-band signals to obtain a full-band noise reduction signal.

[0007] In some possible embodiments of the present disclosure, the provided noise reduction processing method further comprises: receiving a microphone real-time data stream and an acceleration sensor real-time data stream; down-sampling the microphone real-time data stream to obtain the error signal; down-sampling the acceleration sensor real-time data stream to obtain the accelerometer signal.

[0008] In the above embodiments, the analysis amount and the calculation burden in subsequent data processing can be greatly reduced on the premise of retaining useful information, and the noise reduction processing efficiency is improved.

[0009] In some possible embodiments of the present disclosure, the bandwidth and the center frequency of each sub-band signal are determined based on the concentration degree of noise energy in the in-vehicle noise spectrum distribution characteristics.

[0010] In the above embodiments, the key noise components can be more accurately captured, and a more targeted signal basis is provided for subsequent active noise reduction processing.

[0011] In some possible embodiments of the present disclosure, based on the energy analysis of the sub-band signals and the noise scene, the noise reduction processing parameters corresponding to each of the sub-band signals are determined, including: calculating the real-time energy of the sub-band signals, estimating the noise activity, and obtaining an energy analysis result; based on the energy analysis result and the noise scene, dynamically adjusting the noise reduction processing parameters corresponding to each of the sub-band signals; The noise reduction processing parameters at least include at least one or any combination of the following: a noise reduction gain weight, a filter step size, and a processing priority.

[0012] In the above embodiments, not only the changes in the overall noise scene can be responded to, but also the instantaneous dynamics of each frequency band noise can be finely tracked, so that the naturalness, comfort and calculation efficiency of the in-vehicle acoustic environment can be maximized while ensuring the effectiveness of noise reduction.

[0013] In some possible embodiments of the present disclosure, the noise reduction signals corresponding to a plurality of the sub-band signals are fused and reconstructed to obtain a full-band noise reduction signal, including: The noise reduction signals corresponding to a plurality of the sub-band signals are weighted and superimposed according to the noise reduction gain weight to reconstruct the full-band noise reduction signal.

[0014] In some possible embodiments of the present disclosure, based on the noise reduction processing parameters corresponding to each of the sub-band signals, each of the sub-band signals is adaptively filtered in combination with the error signal to obtain a noise reduction signal corresponding to each of the sub-band signals, including: analyze a spectral structure feature of each of the sub-band signals, and select a p-norm based on an analysis result of the spectral structure feature; configure an adaptive filter for each of the sub-band signals based on the noise reduction processing parameter, the selected p-norm, and the error signal, and generate a noise reduction signal corresponding to each of the sub-band signals.

[0015] In some possible embodiments of the present disclosure, a noise scene in which the vehicle is located is identified based on the error signal and the accelerometer signal of the vehicle, and the noise scene in which the vehicle is located is identified based on the error signal and the accelerometer signal of the vehicle. The error signal and the accelerometer signal are input to a pre-trained noise scene identification model for identification, to obtain the noise scene in which the vehicle is located.

[0016] In some possible embodiments of the present disclosure, the noise reduction processing method further includes: performing anomaly detection on the full-band noise reduction signal to obtain an anomaly detection result; In response to the anomaly detection result triggering a noise anomaly, recording noise reduction processing data, a noise scene, and a frequency band distribution of abnormal noise of each of the sub-band signals at present as an updated training sample; retraining the noise scene identification model based on the updated training sample.

[0017] In the above embodiments, the identification capability for a new noise scene can be improved.

[0018] In some possible embodiments of the present disclosure, the noise reduction processing method further includes: updating a sub-band division strategy and a noise reduction processing parameter adjustment rule based on the updated training sample.

[0019] In the above embodiments, an intelligent noise reduction ecology of perception-decision-execution-feedback-evolution can be constructed.

[0020] According to a second aspect of the present disclosure, a noise reduction processing apparatus is provided, including: a scene identification unit configured to identify a noise scene in which a vehicle is located based on an error signal and an accelerometer signal of the vehicle; a signal decomposition unit configured to decompose the accelerometer signal into a plurality of sub-band signals that do not overlap with each other; a parameter determination unit configured to determine, for each of the sub-band signals, a noise reduction processing parameter corresponding to each of the sub-band signals based on energy analysis of the sub-band signal and the noise scene; a filtering unit configured to perform adaptive filtering on each of the sub-band signals based on the noise reduction processing parameter corresponding to each of the sub-band signals and the error signal, to obtain a noise reduction signal corresponding to each of the sub-band signals; The fusion reconstruction unit is configured to fuse and reconstruct the noise reduction signals corresponding to the plurality of sub-band signals to obtain a full-band noise reduction signal.

[0021] In some possible embodiments of the present disclosure, the filtering unit is configured to: analyze a spectral structure feature of each of the sub-band signals, and select a p-norm based on a result of the spectral structure feature analysis; configure an adaptive filter for each of the sub-band signals based on the noise reduction processing parameter and the selected p-norm, and generate a noise reduction signal corresponding to each of the sub-band signals based on the error signal.

[0022] In some possible embodiments of the present disclosure, the scene recognition unit is configured to: input the error signal and the accelerometer signal into a pre-trained noise scene recognition model to obtain a noise scene in which the vehicle is located.

[0023] In some possible embodiments of the present disclosure, the noise reduction device further includes an abnormal feedback unit configured to: perform abnormal detection on the full-band noise reduction signal to obtain an abnormal detection result; in response to the abnormal detection result triggering a noise abnormality, record noise reduction processing data, a noise scene, and a frequency band distribution of the abnormal noise of each of the sub-band signals as an update training sample; retrain the noise scene recognition model based on the update training sample.

[0024] According to a third aspect of the embodiments of the present disclosure, a vehicle is provided, including: a processor; a memory for storing processor-executable instructions; The processor is configured to implement the steps of any one of the noise reduction processing methods of the first aspect.

[0025] According to a fourth aspect of the embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, when instructions in the storage medium are executed by a processor of a terminal, the terminal is enabled to execute any one of the noise reduction processing methods of the first aspect.

[0026] According to a fifth aspect of the embodiments of the present disclosure, a computer program product is provided, including a computer program, when the computer program is executed by a processor, the computer program implements any one of the noise reduction processing methods of the first aspect.

[0027] The technical solutions provided by the embodiments of the present disclosure can include the following beneficial effects: The disclosure identifies the noise scene in which the vehicle is located based on the error signal and the accelerometer signal of the vehicle; decomposes the accelerometer signal into a plurality of mutually non-overlapping sub-band signals; for each sub-band signal, determines the noise reduction processing parameter corresponding to each sub-band signal based on the energy analysis of the sub-band signal and the noise scene; based on the noise reduction processing parameter corresponding to each sub-band signal, combines the error signal to adaptively filter each sub-band signal to obtain the noise reduction signal corresponding to each sub-band signal; and fuses and reconstructs the noise reduction signals corresponding to the plurality of sub-band signals to obtain a full-band noise reduction signal. By decomposing the noise signal to be processed into a plurality of mutually non-overlapping narrow-band sub-signals, and adaptively filtering and reducing each sub-band signal based on the noise scene, the complex and multi-spectrum structure noise is efficiently divided and responded, the noise reduction effect is improved, and the user's driving experience is improved.

[0028] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0029] The accompanying drawings incorporated in the specification and forming a part of it, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure.

[0030] Figure 1 is a flowchart of a noise reduction processing method according to an exemplary embodiment of the disclosure.

[0031] Figure 2 is an implementation flowchart of step S130 according to an exemplary embodiment of the disclosure.

[0032] Figure 3 is an implementation flowchart of step S140 according to an exemplary embodiment of the disclosure.

[0033] Figure 4 is a flowchart of another noise reduction processing method according to an exemplary embodiment of the disclosure.

[0034] Figure 5 is a structural block diagram of a noise reduction processing device according to an exemplary embodiment of the disclosure.

[0035] Figure 6 is a block diagram of a vehicle according to an exemplary embodiment of the disclosure. DETAILED DESCRIPTION

[0036] Exemplary embodiments of the present disclosure will be described herein below with reference to a few examples only, as shown in the accompanying drawings. The following description, in relation to the drawings, is not intended to limit the scope of the present disclosure as set out in the claims. Various changes, modifications, and equivalents thereof will become apparent to those of ordinary skill in the art once the present disclosure is made. For instance, the order of operations described herein is merely examples, and is not limited to those set forth herein, but can be changed as apparent to one of ordinary skill in the art once the present disclosure is made. In addition, the description of features known in the art can be omitted for the sake of clarity and conciseness.

[0037] The implementations described in the following exemplary embodiments of the present disclosure do not represent all implementations consistent with the present disclosure. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0038] The specific implementation of the embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0039] Figure 1 FIG. 1 is a flowchart of a noise reduction processing method according to an exemplary embodiment of the present disclosure, which can be used in a vehicle, which can be a new energy vehicle.

[0040] In some embodiments of the present disclosure, the noise reduction processing method provided includes the following steps: In step S110, the noise scene in which the vehicle is located is identified based on the error signal and the accelerometer signal of the vehicle.

[0041] It should be noted that the error signal is obtained from the sound signal collected by the microphone in the vehicle in real time, which can be the sound signal collected by the microphone arranged inside the cabin, such as near the driver's ear or in the passenger area, reflecting the actual residual noise level currently existing in the vehicle, especially for measuring the noise components that have not been completely cancelled by the active noise reduction system ANC, containing acoustic information directly related to the auditory perception of the crew, such as motor high-frequency whistling, wind noise, road noise, etc.

[0042] It should be noted that the accelerometer signal represents the data stream collected by the acceleration sensor in the vehicle in real time, and the acceleration sensor can be installed at key parts of the vehicle body, such as the suspension, chassis or vehicle body structure, for real-time monitoring of the vibration state of the vehicle structure. These vibrations are often caused by road excitation, engine / motor operation or air dynamic disturbance at high speed, and can be converted into in-vehicle noise, i.e. structure-borne noise, through structural conduction. The accelerometer signal can effectively capture the frequency, amplitude and time-varying characteristics of these mechanical vibrations.

[0043] In some embodiments of the present disclosure, the provided noise reduction processing method further comprises: receiving a microphone real-time data stream and an acceleration sensor real-time data stream; performing down-sampling processing on the microphone real-time data stream to obtain an error signal; and performing down-sampling processing on the acceleration sensor real-time data stream to obtain an accelerometer signal. The original sampling rate of the microphone real-time data stream and the acceleration sensor real-time data stream is relatively high, and the data volume is huge. By performing down-sampling, the analysis amount and the calculation burden in subsequent data processing can be greatly reduced on the premise of retaining useful information, and the noise reduction processing efficiency can be improved.

[0044] In some embodiments of the present disclosure, by analyzing the above two signals, discriminative features such as spectral distribution, energy concentrated frequency band, and time domain correlation can be extracted, and a pre-trained model can be used to determine in which typical noise scene the vehicle is currently in, such as high-speed cruising wind noise dominance, low-frequency road noise caused by rough road surface, electromagnetic howling generated by motor acceleration, etc. The determined noise scene can provide more rich key context information for subsequent adaptive filter parameter adjustment, noise reduction strategy adjustment, or acoustic comfort optimization, etc., so as to realize more accurate and efficient in-vehicle noise control.

[0045] In some embodiments of the present disclosure, the error signal and the accelerometer signal can be input into a pre-trained noise scene recognition model for recognition to obtain the noise scene in which the vehicle is located. It should be noted that the noise scene recognition model can be pre-trained by the vehicle, and the noise scene recognition model can also be pre-trained by the cloud based on an AI model, such as a machine learning based, neural network based, etc. manner, and then downloaded to the vehicle for use by the vehicle.

[0046] In step S120, the accelerometer signal is decomposed into a plurality of mutually non-overlapping sub-band signals.

[0047] It should be noted that the accelerometer signal is a wideband signal, which can contain wideband components from low frequency to medium-high frequency. These components can correspond to different physical noise sources, such as low-frequency road noise, medium-frequency suspension resonance, and high-frequency motor electromagnetic vibration. If the full-band signal is directly used for processing, it is easy to cause feature aliasing, control precision degradation, or waste of computing resources. Sub-band decomposition can achieve frequency domain decoupling and improve the perception and response ability of the vehicle to specific frequency band noise.

[0048] In some embodiments of the present disclosure, the accelerometer signal can be decomposed into a plurality of mutually non-overlapping sub-band signals based on a sub-band division strategy through a digital filter bank, such as a uniform or non-uniform band-pass filter, or a fast Fourier transform (FFT). The original wideband vibration signal is finely processed in the frequency domain to more accurately match the frequency characteristics of different noise sources and provide structured input features for subsequent active noise reduction processing.

[0049] In implementation, a set of digital band-pass filters with non-overlapping center frequencies covering the target analysis band can be set up, and a finite impulse response (FIR) or an infinite impulse response (IIR) structure can be used. Each filter allows signals of a specific frequency band to pass, thereby separating the original accelerometer signal into multiple narrow-band sub-signals, which are continuous in time domain and have controllable delays, and are suitable for real-time embedded system deployment. Alternatively, the accelerometer signal can be converted to the frequency domain through FFT, the frequency spectrum is then divided into several non-overlapping frequency bands according to a preset sub-band division strategy, and then the frequency domain data of each frequency band is zero-padded or windowed and then inverse FFT (IFFT) is performed to restore the corresponding time domain sub-band signals. This method has high calculation efficiency and flexible frequency band division, and is particularly suitable for offline analysis or vehicle-mounted processors with strong computing power.

[0050] In some embodiments of the present disclosure, when the sub-band signals are decomposed, the sub-band signals are not divided with fixed bandwidths or uniform intervals. The sub-band division strategy can be that the bandwidth and center frequency of each sub-band signal are determined based on the concentration degree of noise energy in the in-vehicle noise spectrum distribution characteristics. That is, the bandwidth and center frequency of each sub-band can be dynamically set according to the energy distribution characteristics of the main noise in the frequency domain, for example, a narrower bandwidth can be used in the frequency band with dense noise energy to improve the resolution, and the bandwidth can be appropriately widened in the energy sparse area, so as to improve the calculation efficiency while ensuring the analysis accuracy. The key noise components can be more accurately captured, and a more targeted signal basis is provided for subsequent active noise reduction processing.

[0051] In step S130, for each sub-band signal, the noise reduction processing parameter corresponding to each sub-band signal is determined based on the energy analysis of the sub-band signal and the noise scene.

[0052] It should be noted that the energy analysis result of the sub-band signal is different, the noise scene is different, and the noise reduction processing parameter corresponding to the sub-band signal can be adaptively adjusted, that is, the noise reduction processing parameter is adjusted based on the adjustment rule.

[0053] In step S140, each sub-band signal is adaptively filtered based on the noise reduction processing parameter corresponding to each sub-band signal and the error signal, and the noise reduction signal corresponding to each sub-band signal is obtained.

[0054] In some embodiments of the present disclosure, an adaptive filter corresponding to each sub-band signal can be set based on the noise reduction processing parameter corresponding to each sub-band signal, and the weight is adjusted in combination with the error signal to generate the noise reduction signal corresponding to each sub-band signal.

[0055] In step S150, the noise reduction signals corresponding to the plurality of sub-band signals are fused and reconstructed to obtain a full-band noise reduction signal.

[0056] In some embodiments of the present disclosure, the noise reduction signals corresponding to the plurality of sub-band signals are weighted and superimposed according to the noise reduction gain weights to reconstruct a full-band noise reduction signal for driving the loudspeaker. Different sub-bands can correspond to different noise source characteristics, and therefore the noise reduction strength, i.e. the gain weight, of each sub-band can be independently adjusted to achieve fine control in the frequency domain. For example, when the noise energy is strong and the human ear is sensitive in a certain frequency band, a higher noise reduction gain can be given; while in a frequency band with weak energy or prone to distortion, the gain is appropriately reduced to protect the sound quality.

[0057] In specific implementation, in response to the embodiment of setting a group of digital band-pass filters for splitting, the sub-band noise reduction signals have been generated in the time domain, and can be directly added according to the weights to complete the linear superposition of narrow-band filtering output and obtain a full-band time-domain noise reduction signal. In response to the embodiment of sub-band decomposition based on FFT, after the corresponding noise reduction gain is applied to each sub-band in the frequency domain, the sub-bands can be uniformly weighted and combined in the frequency domain, and then reconstructed into a time-domain full-band noise reduction signal through one IFFT.

[0058] It should be noted that in the sub-band signal fusion process, the continuity and smoothness of the reconstructed signal in amplitude and phase must be ensured. Thus, while achieving efficient wideband noise reduction, the naturalness, stability and high sound quality of the in-vehicle sound field are guaranteed, meeting the stringent requirements of intelligent cockpit for acoustic experience.

[0059] From the above analysis, it can be seen that in the embodiments of the present disclosure, the noise scene in which the vehicle is located is identified based on the error signal and the accelerometer signal of the vehicle; the accelerometer signal is decomposed into a plurality of mutually non-overlapping sub-band signals; for each sub-band signal, the noise reduction processing parameter corresponding to each sub-band signal is determined based on the energy analysis of the sub-band signal and the noise scene; each sub-band signal is adaptively filtered based on the noise reduction processing parameter corresponding to each sub-band signal in combination with the error signal to obtain a noise reduction signal corresponding to each sub-band signal; and the noise reduction signals corresponding to the plurality of sub-band signals are fused and reconstructed to obtain a full-band noise reduction signal. By decomposing the noise signal to be processed into a plurality of mutually non-overlapping narrow-band sub-signals, and adaptively filtering and reducing each sub-band signal based on the noise scene in which it is located, efficient divide-and-conquer and response to complex and multi-spectrum structure noise are achieved, the noise reduction effect is improved, and the user's driving experience is improved.

[0060] In some embodiments of the present disclosure, the implementation process of step S130 is as shown in Figure 2 The implementation process of step S130 is as shown in

[0061] In step S210, the real-time energy of the sub-band signal is calculated, the noise activity is estimated, and the energy analysis result is obtained.

[0062] It should be noted that the real-time energy calculation is performed on each sub-band signal to quantify the activity level of the vibration or noise component in the frequency band. Specifically, by performing short-time energy estimation on the time-domain signal (or its frequency-domain representation) of each sub-band, such as using sliding window mean square value, RMS energy or logarithmic energy method, a time-varying energy sequence is obtained. The energy value reflects the noise intensity in the frequency band corresponding to the current sub-band, and is used to estimate the noise activity, that is, to determine whether there is significant noise in the frequency band and whether it is in a dynamic state. It is used to provide data support for subsequent noise reduction gain weight allocation and filter step size adaptation.

[0063] In step S220, the noise reduction processing parameters corresponding to each sub-band signal are dynamically adjusted based on the energy analysis results and the noise scene.

[0064] It should be noted that the noise reduction processing parameters include at least one or any combination of the following: noise reduction gain weight, filter step size, processing priority. Noise reduction gain weight: controls the strength of the sub-band noise reduction output. For example, when the energy of a certain sub-band is high and belongs to the key frequency band of the current noise scene, the gain is increased to enhance the cancellation effect; otherwise, the gain is reduced to avoid excessive suppression and introduce artificial noise or distortion. Filter step size (learning rate): used in adaptive filtering algorithm, determines the speed of filter coefficient update. In sub-bands with high noise activity and strong signal non-stationarity, the step size can be appropriately increased to speed up convergence; while in stationary or low-energy sub-bands, the step size is reduced to improve stability and reduce residual error fluctuations. Processing priority: in embedded systems with limited computing resources, processing resources can be dynamically allocated according to the importance of the sub-band. For example, high-energy, high-sensitivity frequency bands are given higher priority to ensure their real-time performance and accuracy; low-contribution frequency bands can reduce the sampling rate, simplify the model or delay processing.

[0065] In some embodiments of the present disclosure, the dynamic adjustment described above is driven by both scene awareness and energy awareness, so that optimal noise reduction performance and sound quality balance are achieved under different working conditions. In implementation, one or more energy thresholds, which can include a low energy threshold and a high energy threshold, can be set, and based on this, different processing strategies can be adopted for different subbands: for subbands with energy lower than the low energy threshold, i.e., frequency bands with low noise activity and weak contribution, the noise reduction gain weight can be significantly reduced, or the noise reduction processing for the subband can be temporarily turned off or frozen. This can avoid applying unnecessary anti-phase sound waves to silent or near-silent frequency bands, and can also save valuable on-board computing resources. For subbands with energy higher than the high energy threshold, i.e., dominant frequency bands in the current noise scene, the noise reduction gain weight can be increased to enhance the cancellation effect, the step size of the adaptive filter can be increased to speed up the convergence, and the processing priority of the subband can be increased. For example, a higher CPU scheduling weight, a shorter processing period, or a more complex algorithm model can be assigned to ensure that the key frequency band has sufficient computing resources and real-time response capability.

[0066] In implementation, the above adjustment can also be deeply integrated with the noise scene context. For example, in a scene identified as high-speed wind noise, even if the energy of a high-frequency subband does not reach the absolute high threshold, if it meets the typical spectral characteristics of wind noise, the system can still moderately increase the processing priority of the subband. In a low-speed urban scene, a low-frequency road noise subband can be given a higher noise reduction gain weight even if its energy is moderate.

[0067] Embodiments of the present disclosure not only respond to changes in the overall noise scene, but also finely track the instantaneous dynamics of each frequency band noise, thereby ensuring the effectiveness of noise reduction while maximizing the naturalness, comfort, and computational efficiency of the in-vehicle acoustic environment.

[0068] In some embodiments of the present disclosure, the implementation process of step S140 includes the following steps as shown in Figure 3

[0069] In step S310, the spectral structure characteristics of each subband signal are analyzed, and based on the spectral structure characteristic analysis result, the p-norm is selected.

[0070] In step S320, based on the noise reduction processing parameters and the selected p-norm, an adaptive filter is configured for each subband signal in combination with the error signal, and a noise reduction signal corresponding to each subband signal is generated.

[0071] It should be noted that the spectral structure characteristics of each subband signal are analyzed to deeply depict the inherent statistics and spectral characteristics of the noise. The following key spectral characteristics can be calculated: ​Spectral Flatness: Quantifies the "flatness" of the spectrum by calculating the ratio of the geometric mean and the arithmetic mean of the subband spectrum. A ratio close to 1 indicates a flat signal spectrum, tending to be a wideband noise such as wind noise; a ratio approaching 0 indicates that the energy is concentrated in a few frequencies, typical of narrowband noise such as motor electromagnetic howling.

[0072] Kurtosis: Measures the sharpness of the amplitude distribution of the signal, reflecting its sparsity or impulsiveness. High kurtosis usually corresponds to transient impact or sparse interference, such as structure vibration pulses caused by road bumps, and low kurtosis is more common in stationary Gaussian noise.

[0073] Energy Concentration: Defined as the ratio of the maximum spectral line energy to the total energy, used to determine whether the noise energy is highly concentrated in a specific frequency point, and to assist in identifying resonances or single-frequency interference.

[0074] The above features are normalized, such as Min-Max scaling to the [0, 1] interval, and can be input to a pre-set noise type discrimination module. The module can be based on a rule base, such as if the spectral flatness is <0.3 and the energy concentration is >0.6, it is determined to be narrowband howling. Or a lightweight AI model, such as a decision tree or a small neural network, is used for classification, and the noise spectrum type to which the current subband belongs is output.

[0075] In some embodiments of the present disclosure, the p-norm is selected based on the analysis results of the above spectral structure features, that is, the optimal p-norm estimation is performed, and the spectral flatness and other continuous features are used to dynamically determine the optimal p-norm applicable to the subband through an interpolation function, such as a linear or piecewise nonlinear mapping. For example, when the noise working condition is stable, a steady-state filter is used for the p-norm, such as a first-order low-pass smoothing, to avoid filter oscillation caused by frequent jumps.

[0076] In some embodiments of the present disclosure, based on the noise reduction processing parameters, the estimated optimal p-norm, and the error signal and the filtered reference signal, an adaptive filter with variable step size and p-norm normalization is independently configured for each subband to generate a noise reduction output signal corresponding to each subband signal. The filtered reference signal is the result of filtering each subband signal through a secondary path estimation model, which represents the sound response that the current subband signal will produce at the error microphone when driving the loudspeaker.

[0077] In some embodiments of the present disclosure, the subband signal is: wherein, the index represents the subband signal and takes a value from 1 to N; represents the time.

[0078] The filter output is:

[0079] wherein, is a noise reduction output, characterizes an adaptive weight vector of the subband signal of the reference signal, characterizes a filtered reference signal, , characterizes a convolution operation; characterizes a secondary path impulse response; error calculation is:

[0080] wherein, characterizes an error signal of the subband signal of the reference signal.

[0081] weight update with p-norm normalization:

[0082] wherein, characterizes a p-norm of the subband signal of the reference signal; is a very small number to prevent the denominator from being 0; step size can be adaptively adjusted according to the energy of the subband signal, for example, a logarithmic compression mapping can be used, so as to ensure fast convergence in high-energy subbands and stability in low-energy subbands.

[0083]

[0084] wherein, characterizes a minimum step size, characterizes a maximum step size; characterizes a real-time energy of the subband signal of the reference signal; characterizes a minimum real-time energy; characterizes a maximum real-time energy.

[0085] It should be noted that in the above processing process, the processing of multiple subband signals, including filtering, updating and output generation, are asynchronously and concurrently executed, fully utilizing the parallel computing capability of the vehicle-mounted multi-core processor or DSP, and meeting the real-time requirement.

[0086] In some embodiments of the present disclosure, in order to further ensure the quality of the audio output and the system compatibility, the full-band noise reduction signal is up-sampled to the target playback sampling rate after being generated, and the processed sound signal is transmitted to the loudspeaker for playback to generate an anti-phase sound wave for canceling the noise in the vehicle.

[0087] The embodiments of the present disclosure also provide a noise reduction processing method, as shown in the following. Figure 4 Steps S410 to S450 in the figure correspond to steps S110 to S150 in the method, which will not be repeated here. Based on the implementation process shown in the figure, the following steps are further included. Figure 1 Figure 1

[0088] In step S460, the full-band noise reduction signal is subjected to anomaly detection to obtain an anomaly detection result.

[0089] In step S470, in response to the anomaly detection result triggering noise anomaly, the noise reduction processing data of the current each sub-band signal, the noise scene, and the frequency band distribution of the abnormal noise are recorded as updated training samples.

[0090] In step S480, the noise scene recognition model is retrained based on the updated training samples.

[0091] It should be noted that, based on the implementation process shown in the figure, a self-learning closed-loop mechanism based on abnormal feedback is further introduced to realize the continuous optimization and long-term self-evolution ability of the active noise reduction system. Through online anomaly detection, data recovery, and model retraining, the vehicle-mounted noise reduction system can adapt to new noise types, vehicle state changes, or environmental condition drifts. Figure 1

[0092] In some embodiments of the present disclosure, after the noise reduction signal is fused and reconstructed into a full-band output, real-time anomaly detection is performed on it. The anomaly can be manifested as one or more of the following phenomena: sudden "puff sound", "click sound" or tone distortion; abnormal increase in residual energy: the error signal significantly deviates from the normal range in a certain period or frequency band; noise reduction gain and actual effect do not match: for example, residual error does not decrease under high gain, indicating that the model is mismatched or a new noise source appears; signal amplitude / phase jitter: may be caused by unstable filter or abnormal reference signal.

[0093] In specific implementation, anomaly detection can be achieved by time domain threshold judgment, frequency domain residual spectrum analysis, or a lightweight anomaly classification model.

[0094] After detecting the noise anomaly, the noise reduction processing data of each sub-band signal can be recorded, which can include: real-time energy, p-norm parameter, step size, filter weight, noise reduction gain, etc. of the sub-band. The noise scene label obtained by analysis and the frequency band distribution characteristics of the monitored abnormal noise, such as which sub-band residual energy suddenly increases, spectrum flatness / kurtosis anomaly, etc., are used to locate the frequency domain position and type of the new noise. ​​​

[0095] By labeling the above structured data as abnormal, high-quality update training samples can be formed for subsequent model iteration. The update training samples can be uploaded to the cloud or the vehicle-mounted learning module for incremental training or fine-tuning of the existing noise scene recognition model, so as to improve the recognition ability of new noise scenes.

[0096] It should be noted that the noise scene recognition model is trained and maintained by the cloud, and the optimized noise scene model can be downloaded to the vehicle terminal in the OTA (over the air) upgrade support mode, so as to realize online evolution of the algorithm strategy without user intervention.

[0097] In some embodiments of the present disclosure, the noise reduction processing method provided further includes updating the sub-band division strategy and the noise reduction processing parameter adjustment rule based on the update training sample. It can be understood that when noise anomalies occur, there can be multiple factors, and the sub-band division strategy and the noise reduction processing parameter adjustment rule can also be re-adjusted to ensure the accuracy of noise processing. For example, by analyzing the abnormal noise frequency band distribution in the update training sample, the shortcomings of the existing sub-band division can be identified, and the sub-band center frequency and bandwidth can be automatically adjusted, or the band division type can be switched. The adjustment of the noise reduction processing parameter usually depends on a preset rule or a static mapping function, and these rules can fail when facing new noise or complex coupling conditions. The parameter-performance correlation data of each sub-band in the update training sample before and after the anomaly occurs can be used to optimize the noise reduction processing parameter adjustment rule, such as a parameter mapping function, a gain control logic, or a p-norm selection strategy.

[0098] It should be noted that the improved sub-band division scheme (such as adjusting the center frequency / bandwidth), the updated filter initialization parameter, or the p-norm mapping rule can be downloaded through OTA and take effect at the next system startup or online hot update, thereby improving the frequency domain matching accuracy of subsequent noise reduction processing.

[0099] The embodiments of the present disclosure construct an intelligent noise reduction ecology of perception-decision-execution-feedback-evolution through the above-mentioned closed-loop processing based on self-learning. Not only can known noise be effectively suppressed, but unknown or evolving noise environments can also be actively discovered, learned, and adapted to, ensuring that the intelligent cabin of the new energy vehicle continuously provides high-quality acoustic comfort and auditory experience.

[0100] From the above analysis, the embodiments of the present disclosure can decompose the complex broadband noise in the cabin into multiple sub-bands, specifically process the noise in each frequency band, and have stronger suppression capability and greatly improved overall noise reduction effect. The number of sub-bands, bandwidth, filter parameters, and p-norm can be dynamically adjusted according to the energy and spectral characteristics of the sub-bands, which can adapt to different vehicle speeds, road conditions, and environmental changes, further improving the noise reduction effect. The multi-core parallel architecture is adopted to reasonably allocate sub-band processing tasks, maximize chip utilization, meet the real-time and low-power requirements of the vehicle noise reduction system, avoid sound quality loss caused by excessive or insufficient processing of the full frequency width, ensure smooth transition of sound quality through sub-band fusion, and maintain stable and efficient operation in the long term. The noise reduction system can continuously optimize itself according to actual feedback, adapt to changes in the vehicle environment, and improve noise reduction accuracy.

[0101] It should be noted that the acquisition, storage, use, processing, etc. of information or data in the technical solutions of the present disclosure comply with relevant provisions of national laws and regulations.

[0102] The embodiments of the present disclosure also provide a noise reduction processing device, Figure 5 is a structural block diagram of a noise reduction processing device according to some embodiments of the present disclosure. As Figure 5 indicated, the noise reduction processing device 500 includes a scene recognition unit 501, a signal decomposition unit 502, a parameter determination unit 503, a filtering unit 504, and a fusion reconstruction unit 505.

[0103] The scene recognition unit 501 is configured to recognize the noise scene in which the vehicle is located based on the error signal and the accelerometer signal of the vehicle. The signal decomposition unit 502 is configured to decompose the accelerometer signal into multiple mutually non-overlapping sub-band signals. The parameter determination unit 503 is configured to determine, for each sub-band signal, a noise reduction processing parameter corresponding to each sub-band signal based on energy analysis of the sub-band signal and the noise scene. The filtering unit 504 is configured to perform adaptive filtering on each sub-band signal based on the noise reduction processing parameter corresponding to each sub-band signal and the error signal, to obtain a noise reduction signal corresponding to each sub-band signal. The fusion reconstruction unit 505 is configured to fuse and reconstruct the noise reduction signals corresponding to the multiple sub-band signals to obtain a full-band noise reduction signal.

[0104] It should be noted that the bandwidth and center frequency of each sub-band signal are determined based on the concentration degree of noise energy in the in-vehicle noise spectral distribution characteristics.

[0105] In some exemplary embodiments of the present disclosure, the noise reduction processing device provided further includes a signal acquisition processing unit configured to: receive a microphone real-time data stream and an acceleration sensor real-time data stream; downsample the microphone real-time data stream to obtain an error signal; downsample the acceleration sensor real-time data stream to obtain an accelerometer signal.

[0106] In some example embodiments of the present disclosure, the parameter determination unit 503 is configured to: calculate real-time energy of the sub-band signals, estimate noise activity, and obtain energy analysis results; based on the energy analysis results and the noise scene, dynamically adjust the noise reduction processing parameters corresponding to each sub-band signal.

[0107] It should be noted that the noise reduction processing parameters at least include one or any combination of the following: a noise reduction gain weight, a filter step size, and a processing priority.

[0108] In some example embodiments of the present disclosure, the filter unit 504 is configured to: analyze the spectral structure characteristics of each sub-band signal, and select a p-norm based on the spectral structure characteristic analysis results; based on the noise reduction processing parameters and the selected p-norm, configure an adaptive filter for each sub-band signal in combination with the error signal, and generate a noise reduction signal corresponding to each sub-band signal.

[0109] In some example embodiments of the present disclosure, the fusion reconstruction unit 505 is configured to: weight and superimpose the noise reduction signals corresponding to the plurality of sub-band signals according to the noise reduction gain weight, and reconstruct a full-band noise reduction signal.

[0110] In some example embodiments of the present disclosure, the scene identification unit 501 is configured to input the error signal and the accelerometer signal to a pre-trained noise scene identification model for identification, and obtain a noise scene in which the vehicle is located.

[0111] Correspondingly, the provided noise reduction processing device further includes an abnormality feedback unit configured to: perform abnormality detection on the full-band noise reduction signal to obtain an abnormality detection result; in response to the abnormality detection result triggering a noise abnormality, record the noise reduction processing data of the current each sub-band signal, the noise scene, and the frequency band distribution of the abnormal noise as an updated training sample; retrain the noise scene identification model based on the updated training sample.

[0112] In some embodiments of the present disclosure, the abnormality feedback unit is configured to: Based on the updated training samples, the sub-band division strategy and the noise reduction processing parameter adjustment rule are updated.

[0113] Figure 6 is a block diagram of a vehicle 600 according to an example embodiment. For example, the vehicle 600 can be a hybrid vehicle, or a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles. The vehicle 600 can be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0114] Referring to Figure 6 , the vehicle 600 can include various subsystems, such as an infotainment system 610, a perception system 620, a decision control system 630, a drive system 640, and a computing platform 650. The vehicle 600 can include more or fewer subsystems, and each subsystem can include multiple components. In addition, each subsystem and each component of the vehicle 600 can be interconnected by wired or wireless means.

[0115] In some embodiments, the infotainment system 610 can include a communication system, an entertainment system, a navigation system, and the like.

[0116] The perception system 620 can include several sensors for sensing information of the environment around the vehicle 600. For example, the perception system 620 can include a global positioning system (which can be a GPS system, a Beidou system, or other positioning systems), an inertial measurement unit (IMU), a laser radar, a millimeter wave radar, an ultrasonic radar, and a camera.

[0117] The decision control system 630 can include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0118] The drive system 640 can include components that provide power motion for the vehicle 600. In one embodiment, the drive system 640 can include an engine, an energy source, a transmission system, and wheels. The engine can be one or a combination of an internal combustion engine, an electric motor, an air compression engine. The engine can convert energy provided by the energy source into mechanical energy.

[0119] Part or all of the functions of the vehicle 600 are controlled by the computing platform 650. The computing platform 650 can include at least one processor 651 and a memory 652, and the processor 651 can execute instructions 653 stored in the memory 652.

[0120] The processor 651 can be any conventional processor, such as a commercially available CPU. The processor can also include a graphics processing unit (GPU), a field programmable gate array (FPGA), a system on chip (SOC), an application specific integrated circuit (ASIC), or a combination thereof.

[0121] The memory 652 can be implemented by any type of volatile or nonvolatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic storage, flash memory, magnetic or optical disk.

[0122] In addition to the instructions 653, the memory 652 can also store data, such as road maps, route information, the position, direction, speed, and the like of the vehicle. The data stored in the memory 652 can be used by the computing platform 650.

[0123] In embodiments of the present disclosure, the processor 651 can execute the instructions 653 to complete all or part of the steps of the noise reduction processing method described above.

[0124] In some embodiments of the present disclosure, a non-transitory computer readable storage medium is provided, when the instructions in the storage medium are executed by a processor of a vehicle, the vehicle is enabled to perform the noise reduction processing method described above.

[0125] In some embodiments of the present disclosure, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the noise reduction processing method described above.

[0126] Other embodiments of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the present disclosure cover any and all variations of the present disclosure which come within the scope of the general concept of the disclosure and the non-limiting examples of the present disclosure. The specification and examples are to be considered exemplary only, with the true scope and spirit of the disclosure indicated by the following claims.

[0127] It should be understood that the present disclosure is not limited to the precise structures herein described and illustrated in the drawings, and that various modifications and changes can be made without departing from its scope. The scope of the present disclosure is limited only by the claims that follow.

Claims

1. A noise reduction processing method, characterized in that, include: Based on the vehicle's error signal and accelerometer signal, the noise scene in which the vehicle is located is identified; The accelerometer signal is decomposed into multiple non-overlapping sub-band signals; For each sub-band signal, based on the energy analysis of the sub-band signal and the noise scenario, the noise reduction processing parameters corresponding to each sub-band signal are determined; Based on the noise reduction processing parameters corresponding to each sub-band signal, and combined with the error signal, adaptive filtering is performed on each sub-band signal to obtain the noise reduction signal corresponding to each sub-band signal; The denoised signals corresponding to the multiple sub-band signals are fused and reconstructed to obtain a full-band denoised signal.

2. The noise reduction processing method according to claim 1, characterized in that, Also includes: Receives real-time data streams from the microphone and the accelerometer. The error signal is obtained by downsampling the real-time data stream from the microphone. The accelerometer signal is obtained by downsampling the real-time data stream from the accelerometer.

3. The noise reduction processing method according to claim 1, characterized in that, The bandwidth and center frequency of each sub-band signal are determined based on the degree of concentration of noise energy in the in-vehicle noise spectrum distribution characteristics.

4. The noise reduction processing method according to claim 1, characterized in that, Based on the energy analysis of the sub-band signals and the noise scenario, the noise reduction processing parameters corresponding to each sub-band signal are determined, including: Calculate the real-time energy of the subband signal, estimate the noise activity, and obtain the energy analysis results; Based on the energy analysis results and the noise scenario, the noise reduction processing parameters corresponding to each sub-band signal are dynamically adjusted. The noise reduction processing parameters include at least one of the following or any combination thereof: Noise reduction gain weight, filter step size, and processing priority.

5. The noise reduction processing method according to claim 4, characterized in that, The denoised signals corresponding to multiple sub-band signals are fused and reconstructed to obtain a full-band denoised signal, including: The denoised signals corresponding to the multiple sub-band signals are weighted and superimposed according to the denoising gain weight to reconstruct the full-band denoised signal.

6. The noise reduction processing method according to claim 1, characterized in that, Based on the noise reduction processing parameters corresponding to each sub-band signal, and in conjunction with the error signal, adaptive filtering is performed on each sub-band signal to obtain the noise-reduced signal corresponding to each sub-band signal, including: Analyze the spectral structure characteristics of each sub-band signal, and select the p-norm based on the spectral structure characteristic analysis results; Based on the noise reduction processing parameters and the selected p-norm, and combined with the error signal, an adaptive filter is configured for each sub-band signal to generate a noise reduction signal corresponding to each sub-band signal.

7. The noise reduction processing method according to claim 1, characterized in that, Based on the vehicle's error signal and accelerometer signal, the noise environment in which the vehicle is located is identified, including: The error signal and the accelerometer signal are input into a pre-trained noise scene recognition model for identification, thereby obtaining the noise scene in which the vehicle is located.

8. The noise reduction processing method according to claim 7, characterized in that, Also includes: Anomaly detection is performed on the full-band noise-reduced signal to obtain anomaly detection results; In response to the noise anomaly triggered by the anomaly detection result, the noise reduction processing data, noise scene, and frequency band distribution of the abnormal noise of each sub-band signal are recorded as updated training samples. Based on the updated training samples, the noisy scene recognition model is retrained.

9. The noise reduction processing method according to claim 8, characterized in that, Also includes: Based on the updated training samples, the subband partitioning strategy and noise reduction parameter adjustment rules are updated.

10. A noise reduction processing device, characterized in that, include: The scene recognition unit is used to identify the noise scene in which the vehicle is located based on the vehicle's error signal and accelerometer signal. The signal decomposition unit is used to decompose the accelerometer signal into multiple non-overlapping sub-band signals; The parameter determination unit is used to determine the noise reduction processing parameters corresponding to each sub-band signal based on the energy analysis of the sub-band signal and the noise scene. The filtering unit is used to adaptively filter each sub-band signal based on the noise reduction processing parameters corresponding to each sub-band signal and in combination with the error signal to obtain the noise reduction signal corresponding to each sub-band signal; The fusion and reconstruction unit is used to fuse and reconstruct the noise-reduced signals corresponding to multiple sub-band signals to obtain a full-band noise-reduced signal.

11. The noise reduction processing device according to claim 10, characterized in that, The filtering unit is configured as follows: Analyze the spectral structure characteristics of each sub-band signal, and select the p-norm based on the spectral structure characteristic analysis results; Based on the noise reduction processing parameters and the selected p-norm, and combined with the error signal, an adaptive filter is configured for each sub-band signal to generate a noise reduction signal corresponding to each sub-band signal.

12. The noise reduction processing device according to claim 10, characterized in that, The scene recognition unit is configured as follows: The error signal and the accelerometer signal are input into a pre-trained noise scene recognition model for identification, thereby obtaining the noise scene in which the vehicle is located.

13. The noise reduction processing device according to claim 12, characterized in that, Also includes: An error feedback unit is used for: Anomaly detection is performed on the full-band noise-reduced signal to obtain anomaly detection results; In response to the noise anomaly triggered by the anomaly detection result, the noise reduction processing data, noise scene, and frequency band distribution of the abnormal noise of each sub-band signal are recorded as updated training samples. Based on the updated training samples, the noisy scene recognition model is retrained.

14. A vehicle, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to implement the noise reduction processing method according to any one of claims 1 to 9.

15. A non-transitory computer-readable storage medium, wherein when instructions in the storage medium are executed by a processor of a terminal, the terminal is enabled to perform the steps of a noise reduction processing method according to any one of claims 1 to 9.

16. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the noise reduction processing method as described in any one of claims 1 to 9.

Images