Noise control method, electronic apparatus, storage medium, and drivable device

Abstract

The invention discloses a noise control method, electronic equipment, a storage medium and a drivable device. The noise control method comprises the following steps: determining a target filter coefficient matched with a disturbance noise signal in a cabin under a current working condition in a pre-training filter database; and performing de-noising control according to the target filter coefficient. According to the method, based on the pre-training filter database, the optimal pre-training filter coefficient in the pre-training filter database can be selected as the target filter coefficient according to the disturbance noise signal in the cabin under the current working condition, so that de-noising control is performed, and the problems of relatively large de-noising operation amount and relatively long convergence time are solved; and the requirement on the computing power of the chip is low.

Description

Technical Field

[0001] The present invention relates to the field of noise reduction, and in particular to a noise control method, an electronic device, a computer-readable storage medium and a drivable device. Background Art

[0002] In the related technology, active noise reduction in a small fixed area in the cockpit can be achieved by an adaptive active noise controller combined with an error microphone module, an acceleration sensor module, and a speaker module. However, when reducing noise, it is necessary to calculate the adaptive filter coefficients online in real time, which requires a large amount of calculation and a long convergence time, and places high requirements on chip computing power. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, one purpose of the present invention is to propose a noise control method, which is based on a pre-trained filter database, and selects the optimal pre-trained filter coefficient in the pre-trained filter database as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working condition, thereby solving the problem of large noise elimination calculation amount and long convergence time, and has low requirements on chip computing power.

[0004] A second objective of the present invention is to provide an electronic device.

[0005] A third object of the present invention is to provide a computer-readable storage medium.

[0006] A fourth objective of the present invention is to provide a traversable device.

[0007] In order to solve the above problems, an embodiment of the first aspect of the present invention provides a noise control method, including: determining a target filter coefficient in a pre-trained filter database that matches the disturbance noise signal in the cabin under current working conditions; and performing noise cancellation control according to the target filter coefficient.

[0008] According to the noise control method of the embodiment of the present invention, based on the pre-trained filter database, the optimal pre-trained filter coefficient in the pre-trained filter database is selected as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working conditions, so as to perform noise cancellation control, which solves the problems of large amount of noise cancellation calculation and long convergence time, and has low requirements on chip computing power.

[0009] In some embodiments, the pre-trained filter database includes filter coefficients corresponding to noises in multiple different frequency bands pre-trained offline.

[0010] In some embodiments, determining a target filter coefficient in a pre-trained filter database that matches a disturbance noise signal in a cockpit under current operating conditions includes: obtaining an index value of the target filter coefficient based on the disturbance noise signal; and determining the target filter coefficient from the pre-trained filter database based on the index value.

[0011] In some embodiments, obtaining the index value of the target filter coefficient according to the disturbance noise signal includes: determining a target noise frequency band that matches the disturbance noise signal, and obtaining the index value of the filter coefficient in the pre-trained filter database corresponding to the target noise frequency band as the index value of the target filter coefficient.

[0012] In some embodiments, a target noise frequency band matching the disturbance noise signal is determined, and an index value of a filter coefficient in the pre-trained filter database corresponding to the target noise frequency band is obtained, including: performing frequency band classification on the disturbance noise signal through a frequency band classification neural network model to determine the target noise frequency band and outputting the index value of the target filter coefficient.

[0013] In some embodiments, the frequency band classification neural network model includes a one-dimensional convolutional neural network, which takes the disturbance noise signal as input and takes the index value of the target filter coefficient as output.

[0014] In some embodiments, the one-dimensional convolutional neural network is configured with a residual block.

[0015] In some embodiments, the frequency band classification neural network model includes a two-dimensional convolutional neural network, which takes the disturbance noise signal as input and takes the index value of the target filter coefficient as output.

[0016] In some embodiments, noise cancellation control is performed according to the target filter coefficient, including: controlling a noise cancellation speaker to play according to the target filter coefficient, so that the sound signal played by the noise cancellation speaker reaches the disturbance noise signal collection point and cancels out the disturbance noise signal.

[0017] In some embodiments, the signal size of the sound signal played by the noise cancellation speaker arriving at the disturbance noise signal collection point is expressed as the product of the external noise reference signal, the target filter coefficient and the sound signal played by the noise cancellation speaker.

[0018] In some embodiments, the noise control method further includes: determining the external noise reference signal according to the device speed under the current working condition.

[0019] In some embodiments, the external noise reference signal is determined based on the device speed under the current operating condition, including: when the device speed is greater than a speed threshold, the external noise reference signal is a wind noise signal collected by a wind noise collection device; or, when the device speed is less than or equal to the speed threshold, the external noise reference signal is a road noise signal collected by a road noise collection device.

[0020] In some embodiments, the noise control method further includes: determining the external noise reference signal according to a correlation coefficient between the external noise signal and the disturbance noise signal.

[0021] In some embodiments, the external noise signal includes a road noise signal collected by a road noise collection device and a wind noise signal collected by a wind noise collection device; the external noise reference signal is determined according to the correlation coefficient between the external noise signal and the disturbance noise signal, including: when the correlation coefficient between the road noise signal and the disturbance noise signal is greater than the correlation coefficient between the wind noise signal and the disturbance noise signal, the external noise reference signal is the road noise signal; or, when the correlation coefficient between the wind noise signal and the disturbance noise signal is greater than the correlation coefficient between the road noise signal and the disturbance noise signal, the external noise reference signal is the wind noise signal.

[0022] A second aspect of the present invention provides an electronic device, comprising: at least one processor; a memory communicatively connected to the at least one processor; the memory storing a computer program executable by the at least one processor, wherein the at least one processor implements the noise control method described in the above embodiment when executing the computer program.

[0023] According to the electronic device of the embodiment of the present invention, the corresponding noise control program can be stored in the memory. When implementing the charging control method, the processor runs the program in the memory, obtains the target filter coefficient, and performs noise cancellation control according to the target filter coefficient.

[0024] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the noise control method described in the above embodiment when executed.

[0025] A fourth aspect of the present invention provides a traversable device, which includes the electronic device described in the above embodiment; or, the traversable device includes: a first collection device, which is arranged at a position corresponding to the human ear in the cabin, and is used to collect disturbance noise signals in the cabin; a second collection device, which is used to collect external noise signals; a noise cancellation speaker and a controller, wherein the controller is connected to the first collection device, the second collection device and the noise cancellation speaker, and is used to control the noise cancellation speaker according to the noise control method described in the above embodiment.

[0026] According to the traversable device of the embodiment of the present invention, filter coefficients of multiple noises in different frequency bands are pre-trained offline in advance and stored, and the optimal pre-trained filter coefficient in the pre-trained filter database is selected as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working condition, so as to control the noise cancellation speaker to perform noise cancellation, reduce the amount of noise cancellation calculation and convergence time, and reduce the chip computing power.

[0027] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 is a flow chart of a noise control method according to an embodiment of the present invention; Figure 2 is a schematic diagram of a one-dimensional convolutional neural network according to an embodiment of the present invention; Figure 3 is a schematic diagram of a two-dimensional convolutional neural network according to an embodiment of the present invention; Figure 4 is a schematic diagram of a vehicle active noise control system structure according to an embodiment of the present invention; Figure 5 is a schematic diagram of noise control algorithm logic according to one embodiment of the present invention; Figure 6 is a time-frequency diagram of the noise at the passenger's ear at 110 km / h according to one embodiment of the present invention; Figure 7 This is a diagram showing the effect of active noise control using the FxLMS algorithm according to an embodiment of the present invention (the left side shows the time domain, and the right side shows the sound pressure); Figure 8 is a diagram showing the effect of active noise control by selecting a fixed filter according to an embodiment of the present invention (the left side is the time domain, the right side is the sound pressure); Fig. 9is a diagram showing the effect of active noise control by selecting a fixed filter according to an embodiment of the present invention (the left side is the time domain, the right side is the sound pressure); Fig.10 is a structural block diagram of an electronic device according to an embodiment of the present invention; Fig.11 is a structural block diagram of a drivable device according to an embodiment of the present invention.

[0029] Reference numerals: A movable device 200; an electronic device 100; First collection device 201; second collection device 202; noise canceling speaker 203; controller 204; processor 205; memory 206. DETAILED DESCRIPTION Embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. Embodiments of the present invention are described in detail below.

[0030] Existing noise control methods use vibration sensors, error microphones, noise controllers, and speakers to use the FxLMS (Filtered-X Least Mean Squares) algorithm to complete the adaptive filtering process and achieve road noise reduction. This method has a fast convergence speed and the control parameters can be continuously updated, but it requires the construction of a longer FIR (Finite Impulse Response, non-recursive) filter to identify the secondary channel and adjust the adaptive filter coefficient group in real time. The calculation amount is large and the convergence time is long. In addition, this technology is mainly aimed at conditions where road noise dominates during driving, and does not discuss conditions where wind noise occurs.

[0031] In order to solve the above problems, an embodiment of the first aspect of the present invention provides a noise control method, which can pre-train and store filter coefficients of multiple noises in different frequency bands offline in advance, and select the optimal pre-trained filter coefficient in the pre-trained filter database as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working conditions, so as to perform noise cancellation control, solve the problems of large noise cancellation calculation amount and long convergence time, and have low requirements on chip computing power.

[0032] Reference below Figure 1 A charging control method according to an embodiment of the first aspect of the present invention is described as follows: Figure 1 As shown, the method at least includes step S1 to step S2.

[0033] Step S1, determining a target filter coefficient in a pre-trained filter database that matches a disturbance noise signal in the cabin under current working conditions.

[0034] Specifically, when performing noise control, offline pre-training can be performed first. The offline pre-training stage can be performed using a training data set during the factory delivery process of the movable device, or in subsequent use, noise data under common road conditions or various working conditions can be collected as a training data set for training to obtain the optimal filter coefficients corresponding to noise signals in different frequency bands, and then the obtained filter coefficients are stored in a pre-trained filter database; the movable device can be a vehicle, a ship, an airplane, etc. When the movable device is in a driving state, the disturbance noise signal in the cabin is obtained in real time, and the obtained disturbance noise signal in the cabin is compared with the data in the pre-trained filter database to determine the target filter coefficient in the pre-trained filter database that matches the disturbance noise signal in the cabin under the current working condition. The target filter coefficient can be understood as the coefficient that the filter corresponds to at this time in order to eliminate noise.

[0035] Step S2, performing noise reduction control according to the target filter coefficient.

[0036] Specifically, since the disturbance noise signals in the cockpit obtained by the drivable device under different driving conditions are different, the disturbance noise signals in the cockpit are obtained in real time, and the obtained disturbance noise signals in the cockpit are compared with the data in the pre-trained filter database, and the target filter coefficients in the pre-trained filter database that match the disturbance noise signals in the cockpit under the current conditions are determined. After obtaining the target filter system, noise cancellation control is performed according to the target filter coefficients, thereby reducing the noise in the cockpit.

[0037] According to the noise control method of the embodiment of the present invention, based on the pre-trained filter database, it is only necessary to obtain the disturbance noise signal in the cockpit under the current working condition in real time, and then the optimal pre-trained filter coefficient in the pre-trained filter database can be selected as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working condition, so as to perform noise cancellation control, which solves the problems of large noise cancellation calculation amount and long convergence time, and has low requirements on chip computing power.

[0038] In some embodiments, the pre-trained filter database includes filter coefficients corresponding to noises of multiple different frequency bands pre-trained offline. The filter coefficients in the training filter database can be pre-trained based on training data, that is, the optimal filter coefficients corresponding to each disturbance noise signal are obtained and pre-stored in the controller or storage medium. When performing noise reduction processing, the optimal target filter coefficient can be selected from the training filter database based on the disturbance noise signal under the current working condition to obtain the optimal noise reduction effect. The controller has a small amount of calculation, there is no problem of slow convergence, and the chip computing power requirement is also low.

[0039] In some embodiments, determining a target filter coefficient in a pre-trained filter database that matches a disturbance noise signal in a cockpit under current operating conditions includes: obtaining an index value of the target filter coefficient based on the disturbance noise signal; and determining the target filter coefficient from the pre-trained filter database based on the index value.

[0040] Specifically, the index value in programming usually refers to the database index. The database index is a separate, physical storage structure that sorts the values ​​of one or more columns in a database table. The main purpose of the pre-trained filter database index is to improve the noise reduction efficiency. Since there is a large amount of data in the pre-trained filter database, if a sequential search is used, the search efficiency will be very low. Therefore, the index value of the target filter coefficient is obtained by disturbing the noise signal, and the target filter coefficient is determined from the pre-trained filter database according to the index value, and the position of the target filter coefficient is quickly located, thereby improving the search speed.

[0041] In some embodiments, obtaining an index value of a target filter coefficient according to a disturbance noise signal includes: determining a target noise frequency band that matches the disturbance noise signal, and obtaining an index value of a filter coefficient in a pre-trained filter database corresponding to the target noise frequency band as an index value of the target filter coefficient.

[0042] Specifically, under offline training, the optimal filter coefficients of the optimal control filters in multiple frequency bands are obtained as convergently as possible without computing force cost, the filter coefficients of multiple noise frequency bands pre-trained in advance are stored, and after obtaining the disturbance noise signal in the cockpit, the current noise frequency band is matched with the noise frequency band in the pre-trained filter database to obtain the target noise frequency band, and the index value of the filter coefficient in the corresponding pre-trained filter database is obtained according to the target noise frequency band, and the target filter coefficient is determined from the pre-trained filter database according to the index value.

[0043] In some embodiments, a target noise frequency band matching the disturbance noise signal is determined, and an index value of a filter coefficient in a pre-trained filter database corresponding to the target noise frequency band is obtained, including: performing frequency band classification on the disturbance noise signal through a frequency band classification neural network model to determine the target noise frequency band and outputting the index value of the target filter coefficient.

[0044] Specifically, after obtaining the disturbance noise signal, the disturbance noise signal is sent to the frequency band classification neural network model. The frequency band classification neural network model classifies the noise of the current working condition according to the characteristics of the input signal, that is, which frequency band noise in the current disturbance signal and the pre-trained filter database is most matched, and the target noise band is obtained. Then, the index value of the target filter coefficient is obtained according to the index value of the filter coefficient corresponding to the target noise band in the pre-trained filter database, and the index value of the target filter coefficient is output at the same time.

[0045] In some embodiments, the frequency band classification neural network model includes a one-dimensional convolutional neural network, which takes a disturbance noise signal as input and an index value of a target filter coefficient as output.

[0046] Specifically, a one-dimensional convolutional neural network is a convolutional neural network specifically used to process one-dimensional data; the basic structure includes the following layers: input layer, convolution layer, activation function layer, pooling layer, fully connected layer and output layer.

[0047] The input layer is used to receive one-dimensional sequence data, such as time series, audio signals, etc.; the convolution layer uses a one-dimensional convolution kernel (filter) to slide on the input data to extract local features; the activation function of the activation function layer is used to introduce nonlinearity and enhance the expressiveness of the model; the pooling layer uses pooling operations (such as maximum pooling and average pooling) to reduce the dimension of the data, reduce the amount of calculation, and retain important features; the fully connected layer flattens the output of the previous layer and then connects to the fully connected layer for classification or regression tasks, usually located in the last few layers of the network, for extracting global features and making decisions; the output layer is used to output data. The one-dimensional convolutional neural network of the present invention takes a perturbation noise signal as input, inputs the perturbation noise signal to the one-dimensional convolutional neural network input layer, and after calculation, the one-dimensional convolutional neural network output layer outputs the index value of the target filter coefficient.

[0048] Figure 2 is a schematic diagram of a one-dimensional convolutional neural network according to an embodiment of the present invention, such as Figure 2 As shown in the figure, the one-dimensional convolutional neural network includes a convolution layer, a maximum pooling layer, a one-dimensional BtchNorm layer, a Softmax layer, a fully connected layer, etc.

[0049] In some embodiments, Figure 2 As shown, the one-dimensional convolutional neural network of the embodiment of the present invention is also configured with a residual block.

[0050] Specifically, Figure 2 As shown, the one-dimensional convolutional neural network directly inputs a frame of time-domain normalized noise signal, where one frame is 1 second. In order to train the one-dimensional convolutional neural network frequency selection classification model, the present invention considers generating 100,000 white noises in the frequency range of [20, 4000] Hz, with a sampling rate of 16,000 Hz, which are sent to the one-dimensional convolutional neural network for training as a training set, and 2,000 pieces are generated as a validation set and a test set respectively. A residual block is added to the network to increase the model depth and alleviate the gradient vanishing problem, so that the model has stronger expression and generalization capabilities.

[0051] In some embodiments, the frequency band classification neural network model may adopt a two-dimensional convolutional neural network, which takes a disturbance noise signal as input and an index value of a target filter coefficient as output.

[0052] Specifically, the two-dimensional convolutional neural network shares parameters. When the convolution kernel slides on the input data, the parameters are shared, which greatly reduces the number of parameters of the model. Each output element in the convolution layer is only connected to a part of the input data area, which helps to capture local features. The convolutional neural network has a certain invariance to the translation of the input data. The two-dimensional convolutional neural network of the present invention takes the disturbance noise signal as input, inputs the disturbance noise signal to the two-dimensional convolutional neural network input layer, and after calculation, the two-dimensional convolutional neural network output layer outputs the index value of the target filter coefficient.

[0053] The structure of a two-dimensional convolutional neural network is as follows Figure 3 As shown, the input signal is converted into the time-frequency domain through short-time Fourier transform, and the classification task is performed by extracting the time-frequency features on the image. Figure 2 and Figure 3 The noise signals used in the test are all mixed noise data collected in the cockpit using a microphone next to the co-pilot's ear at a real speed of 110 km / h.

[0054] For example, the noise control method of the present invention is applicable to a movable device, which may be a vehicle, a ship, an airplane, etc. The following description takes a vehicle as an example.

[0055] The device and method of the present invention can realize active noise control of different broadband noises under various operating speeds during vehicle driving, and can use the method of pre-training adaptive filter coefficients to calculate the similarity between the dominant component of the current noise frequency and the data in the pre-trained filter database using an error microphone, and select the closest optimal pre-trained filter coefficient to reduce the current noise. This method avoids the selection of reference signals and the real-time adaptive update of filter coefficients with high computational complexity, and can better solve the problems of large amount of computation and long convergence time for time-varying non-stationary wind noise reduction.

[0056] like Figure 4As shown, when the vehicle is in motion, the tires interact with the ground to generate vibrations, which are transmitted through the body structure and penetrate into the vehicle, forming road noise; while the external airflow interacts with the body structure and gaps to generate noise, forming wind noise. The present invention is considered to be divided into two stages. Stage one is a pre-training stage. In the pre-training stage, the training data set can be used for training during the vehicle leaving the factory, or the noise data of common road conditions and working conditions can be collected during the subsequent use by consumers as a training data set for training. The main principle is to first use the FxLMS algorithm to perform offline pre-training on a series of noises in different frequency bands to obtain the optimal active noise control filter coefficients corresponding to the noise signal of each frequency band, and then store the obtained filter coefficients in a pre-training filter database of a vehicle active noise controller; Stage two is a noise reduction stage. When the vehicle is in a driving state, an error microphone (at the human ear) is used to collect the noise in the vehicle at this time, and the collected noise is sent to the central processing unit and graphics processing unit of the vehicle. The graphics processing unit is used to train a one-dimensional or two-dimensional neural network to select the frequency of the collected noise according to its amplitude, frequency component, time-frequency characteristics, etc., so that a most suitable control filter index value can be output for the current signal, and then the index value is sent to the active noise reduction controller through the CAN (Controller Area Network) bus in the vehicle, so as to select the corresponding adaptive coefficient in the pre-training filter database to reduce the current noise.

[0057] In some embodiments, noise cancellation control is performed according to the target filter coefficient, including: controlling the noise cancellation speaker to play according to the target filter coefficient, so that the sound signal played by the noise cancellation speaker reaches the disturbance noise signal collection point and cancels out the disturbance noise signal.

[0058] Specifically, the disturbance noise signal is offset by controlling the noise cancellation speaker to produce a sound signal played by the noise cancellation speaker with a phase opposite to that of the noise signal and an amplitude equal to that of the noise signal. The target filter coefficient plays a vital role in noise reduction and determines the frequency response and phase characteristics of the sound signal played by the noise cancellation speaker. When noise cancellation control is performed according to the target filter coefficient, it is necessary to use sensors such as microphones to collect disturbance noise signals, which will be used as reference signals. Based on the obtained target filter coefficient and the disturbance noise signal, an anti-noise signal with a phase opposite to that of the disturbance noise signal and an amplitude equal to that of the disturbance noise signal is generated. Finally, the generated anti-noise signal is played through the noise cancellation speaker. When these signals reach the disturbance noise signal collection point, they will offset the original noise signal, thereby achieving a noise cancellation effect.

[0059] In some embodiments, the signal size of the sound signal played by the noise cancellation speaker arriving at the disturbance noise signal collection point is expressed as the product of the external noise reference signal, the target filter coefficient and the sound signal played by the noise cancellation speaker.

[0060] Specifically, under ideal circumstances, when the sound signal played by the noise-canceling speaker reaches the microphone position, it can perfectly predict and generate a sound signal that is completely opposite to the noise signal, then the total noise level received by the microphone will be close to zero. However, in actual applications, due to the influence of various factors, the noise cancellation effect may not reach the ideal state. Therefore, the signal of the sound signal played by the noise-canceling speaker arriving at the disturbance noise signal collection point is expressed as the product of the external noise reference signal, the target filter coefficient and the sound signal played by the noise-canceling speaker.

[0061] For example, in the offline pre-training stage, two algorithms can be used to pre-train the control filter coefficients and form a pre-trained filter database, which are described below respectively.

[0062] The first one is the adaptive active noise control FxLMS algorithm. Specifically, it is assumed that is the disturbance sound pressure signal from the primary sound source (noise source) to the human ear (error microphone), represents the control filter coefficient, represents external reference signals such as wind noise signals, road noise signals, etc. To eliminate noise, the difference signal e As close to 0 as possible, then the error signal It can be expressed as: ; (1) In formula (1), s(n) refers to the impulse response of the secondary path. In the embodiment, s(n) may correspond to the sound signal played by the noise canceling speaker, and n is the nth time. It can be expressed as the signal size of the sound signal played by the noise-canceling speaker reaching the disturbance noise signal collection point.

[0063] The weight (filter coefficient) update equation of the adaptive control filter can be expressed as: ; (2) in is the learning rate (step size), is the estimated signal corresponding to the secondary path.

[0064] The second is the Kalman filter algorithm based on state estimation and observation. It estimates the state of the dynamic system through a series of observation data containing statistical noise. Here, the filter coefficient w(n) is considered as the state equation, and the Kalman recursive equation group for active noise control can be written as: Prediction of new status: ; (3) in, is the estimated value of w(n).

[0065] Prediction of the autocorrelation matrix of the state error: ; (4) is the state error, is the estimated value of the state error.

[0066] Calculation of the Kalman gain matrix: ; (5) Here q(n) is the variance of the observation error, is the Kalman gain, is the filtered external reference signal.

[0067] Calculation of actual status: ; (6) Calculation of the autocorrelation matrix of the actual state error: (7) By using the Kalman recursive equation from formula (2) to (7), the next state value is estimated from the previous state, and finally an ideal adaptive control filter weight can be obtained.

[0068] from Figure 4 As can be seen in the figure, during the vehicle's driving, the external noise data is collected by the reference sensor (accelerometer or patch microphone) as shown by the black dot to obtain the corresponding reference signal , and the error microphone placed in the car at the corresponding position of the human ear (as shown by the white dot) collects the corresponding disturbance signal By using the FxLMS or Kalman Filter algorithm introduced above, a set of adaptive filter coefficients corresponding to the current working condition noise will be obtained, and this set of coefficients will be stored in the database.

[0069] According to the main frequency bands of wind noise and road noise collected in actual experiments, a series of representative white noises from 20-4000 Hz are generated, where the frequency range is selected as [20, 800] Hz, [200, 1000] Hz, [400, 1200] Hz, etc., and the frequency is increased by 200 Hz to cover the noise range under most working conditions. In addition, considering the working conditions of mixed noises, white noises with a wider frequency range such as [20, 4000] Hz, [20, 2000] Hz, and [2000, 4000] Hz are also selected to be generated. These 20 groups of frequency bands are used as the input of the pre-trained filter, and the Least Mean Square (LMS) algorithm is actually used offline on the vehicle to estimate the corresponding secondary path , and then use the FxLMS or Kalman Filter algorithm to perform active noise control. If the FxLMS algorithm is used, the step size (mu = 0.005) and the control filter length (w_len = 1024) are selected. If the Kalman algorithm is used, the filter length (w_len = 256) is used. Finally, a set of filter coefficients corresponding to the optimal convergence of each frequency band can be obtained, and the pre-trained filter coefficients are stored in a database of a vehicle active noise controller.

[0070] In the online noise elimination stage, the disturbance signal d(n) of the current working condition of the vehicle collected by the reference microphone is sent to a neural network on the graphics processor of the vehicle host. This neural network performs a classification task on the noise of the current working condition based on the characteristics of the input signal, that is, which frequency band noise in the pre-trained filter best matches the current disturbance signal d(n). Then, the index of the filter coefficient corresponding to the most matching frequency band noise in the database is returned. This index is transmitted to the vehicle active noise reduction controller through the vehicle CAN bus, and the corresponding adaptive filter coefficient is taken out to actively control the noise collected by the current error microphone.

[0071] After obtaining the optimal control filter coefficient corresponding to the current noise, the noise is controlled using formula (1) in the active noise controller of the vehicle.

[0072] The present invention uses the vehicle CPU and GPU, and the adaptive active noise reduction controller combines the error microphone module, the acceleration sensor module, and the secondary speaker module to perform active noise control in the cabin in a small fixed area. The optimal filter coefficients of multiple noise frequency bands are pre-trained in advance. Under offline training, the optimal control filter of the current frequency band can be converged as well as possible without computing power cost, avoiding the high computing power requirements of real-time adaptive iteration; the optimal active noise control effect is achieved in a very short time (first frame) without the need for a long time of convergence; for non-stationary time-varying signals, such as wind noise, there is a good convergence effect, and for noise with a wider frequency band, as long as the pre-training filter data set is reasonably selected, good noise reduction performance can be obtained.

[0073] In some embodiments, the noise control method further includes: determining an external noise reference signal according to a device speed under current working conditions.

[0074] Specifically, the external noise reference signal may correspond to x(n) in formula (1), wherein the external noise reference signal includes a road noise signal corresponding to the acceleration signal collected by the acceleration sensor, a corresponding wind noise signal collected by a reference microphone such as a patch microphone, and other signals that cause noise to human ears in the cabin.

[0075] In some embodiments, the external noise reference signal is determined according to the device speed under the current working condition, including: when the device speed is greater than the speed threshold, the external noise reference signal is a wind noise signal collected by a wind noise collection device; or, when the device speed is less than or equal to the speed threshold, the external noise reference signal is a road noise signal collected by a road noise collection device.

[0076] Specifically, the drivable device is affected by wind noise and road noise at the same time during driving. The effects of wind noise and road noise on the drivable device are different at different speeds. When the device speed is greater than the speed threshold, the external noise reference signal is the wind noise signal collected by the wind noise collection device. The speed threshold may be 110 km / h. When the device speed is greater than 110 km / h, the noise generated by the wind noise has a greater impact, and the wind noise collection device collects the wind noise signal as the external noise reference signal. When the device speed is less than 110 km / h, the noise generated by the road noise has a greater impact, and the road noise collection device collects the road noise signal as the external noise reference signal.

[0077] In some embodiments, the noise control method further includes: determining an external noise reference signal according to a correlation coefficient between the external noise signal and the disturbance noise signal.

[0078] Specifically, the correlation coefficient is related to the path of the external noise signal propagating to the disturbance noise signal collection point. The correlation coefficient can be obtained by a multiple correlation coefficient calculation method or calculation function. For example, there is a multiple correlation coefficient calculation correlation function in MATLAB.

[0079] In some embodiments, the external noise signal includes a road noise signal collected by a road noise collection device and a wind noise signal collected by a wind noise collection device; the external noise reference signal is determined according to the correlation coefficient between the external noise signal and the disturbance noise signal, including: when the correlation coefficient between the road noise signal and the disturbance noise signal is greater than the correlation coefficient between the wind noise signal and the disturbance noise signal, the external noise reference signal is the road noise signal; or, when the correlation coefficient between the wind noise signal and the disturbance noise signal is greater than the correlation coefficient between the road noise signal and the disturbance noise signal, the external noise reference signal is the wind noise signal.

[0080] Specifically, the acquired external noise signals include road noise signals and wind noise signals, the road noise signal is acquired by a road noise acquisition device, which may be an acceleration sensor, and the wind noise signal is acquired by a wind noise acquisition device, which may be a patch microphone; after acquiring the external noise signal, the correlation coefficient between the external noise signal and the disturbance noise signal is compared to obtain an external noise reference signal; if the correlation coefficient between the road noise signal and the disturbance noise signal is greater than the correlation coefficient between the wind noise signal and the disturbance noise signal, the road noise signal is used as the external noise reference signal; if the correlation coefficient between the wind noise signal and the disturbance noise signal is greater than the correlation coefficient between the road noise signal and the disturbance noise signal, the wind noise signal is used as the external noise reference signal.

[0081] For example, taking a vehicle as an example, when the road noise collection device collects the road noise signal as the external noise reference signal, Figure 4 The vehicle shown includes the following modules: an error microphone module for collecting vehicle cabin noise signals; a secondary speaker module for playing noise cancellation signals; the vehicle's host CPU (Central Processing Unit) / GPU (Graphics Processing Unit) module; an active noise reduction controller module that collects reference signals and remote microphone signals and generates noise cancellation signals by selecting the optimal filter in the pre-trained filter database and drives the speaker to sound; an error microphone, vehicle host CPU / GPU, active noise reduction controller module, and secondary speaker module are used to actively reduce noise near the human ear to achieve noise reduction, wherein the secondary speaker module and error microphone module are placed on the seat headrest. The logic of the active noise control method is as follows: Figure 5 As shown: Step 1: Select a suitable pre-training dataset for vehicle driving conditions.

[0082] Step 2: Select the appropriate step size, running time and filter order, and use the FxLMS or Kalman Filter algorithm to train the pre-training data set offline to obtain the corresponding optimal filter coefficients.

[0083] Step 3, storing the filter coefficients in the database of the active noise controller module.

[0084] Step 4: When the vehicle speed reaches a certain speed, the active noise reduction module is turned on, the microphone at the error point starts to collect the current noise, and the noise is sent to the vehicle's CPU. The pre-trained convolutional neural network model is used for inference to obtain the filter coefficient index that best matches the current noise. According to this index, the optimal filter coefficient is taken out from the storage of the active noise control system.

[0085] Step 5, update the matched filter in step 4 to the active noise control system parameters, and then use this filter coefficient to control the noise.

[0086] When the vehicle speed is greater than 110 km / h, the mixed noise of wind noise and road noise is collected at the position of the co-pilot's ear. At this time, the wind noise is dominant, and its time-frequency diagram is as follows Figure 6 As shown in the figure, the noise reduction effects obtained by the traditional FxLMS algorithm and the fixed selection filter algorithm are shown in the figure. Figure 7 , Figure 8 , Fig. 9 As shown, the filter order adopted by the FxLMS algorithm is 256, and the step size mu is 0.005. The filter order adopted by the fixed filter algorithm in the pre-training stage is 1024, and the step size is 0.0005. The noise signal before denoising is 63.85dB. The traditional FxLMS algorithm needs to run for 50.9s before the denoising effect can reach about 50.01dB after convergence. The fixed selection filter algorithm based on FxLMS pre-training can reach the optimal filter effect of 27.78dB after 9.2s of running time. The fixed selection filter algorithm based on Kalman Filter pre-training can reach a denoising effect of 23.68dB after running for 9.0s. The denoising time is short and the denoising performance is good, and no real-time convergence is required, which illustrates the advancement of the present invention.

[0087] The present invention uses the vehicle CPU and GPU, and the adaptive active noise reduction controller combines the error microphone module, the acceleration sensor module, and the secondary speaker module to perform active noise control in the vehicle cabin in a small fixed area. A deep learning network is used to match the pre-trained filter and the current noise, so that the optimal adaptive filter coefficient closest to the current noise can be selected in the pre-trained filter library. The optimal noise reduction effect can be achieved in a very short time, and the long learning and convergence in the traditional adaptive FxLMS algorithm are not required. Passengers can have a better noise reduction experience and can quickly feel the effect after turning on the noise reduction function, which improves the quietness of the cabin; at the same time, the computing power requirements of the active noise control system are greatly reduced, and the performance can be achieved with the FxLMS algorithm that converges for a long enough time or even better. Its noise reduction performance mainly depends on the selection of the frequency range of the pre-trained filter and the noise reduction performance of the pre-trained filter. If a longer signal and a sufficiently small step size are used in the pre-training stage, the ideal noise reduction effect can be achieved in theory; in addition, the classification accuracy of the neural network determines whether the most suitable filter coefficient can be selected for the current noise.

[0088] A second aspect of the present invention provides an electronic device, such as Fig.10As shown, the electronic device 100 includes: at least one processor 101 and a memory 102 .

[0089] Among them, at least one processor 101 is communicatively connected with the memory 102, and the memory 102 stores a computer program that can be executed by the at least one processor 101. When the at least one processor 101 executes the computer program, the noise control method is implemented.

[0090] According to the electronic device of the embodiment of the present invention, the corresponding noise control program can be stored in the memory. When implementing the charging control method, the processor runs the program in the memory, obtains the target filter coefficient, and performs noise cancellation control according to the target filter coefficient.

[0091] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed to provide a noise control method.

[0092] A fourth aspect of the present invention provides a drivable device, such as Fig.11 As shown, the traversable device 200 includes the electronic device 100 , or the traversable device 200 includes: a first collection device 201 , a second collection device 202 , a noise cancellation speaker 203 and a controller 204 .

[0093] The first acquisition device 201 is disposed at a position corresponding to a human ear in the cabin and is used to collect disturbance noise signals in the cabin, for example Figure 4 The device shown includes the error microphone mentioned above; the second acquisition device 202 is used to collect external noise signals, such as the acceleration sensor acquisition signal mentioned above and / or the wind noise and other signals collected by the reference microphone such as the patch microphone; the controller 204 is connected to the first acquisition device 201, the second acquisition device 202 and the noise cancellation speaker 203, and is used to control the noise cancellation speaker 203 according to the noise control method of the above embodiment. Among them, the noise cancellation speaker 203 can include the secondary speaker mentioned above such as Figure 4 shown.

[0094] According to the traversable device of the embodiment of the present invention, filter coefficients of multiple noises in different frequency bands are pre-trained offline in advance and stored, and the optimal pre-trained filter coefficient in the pre-trained filter database is selected as the target filter coefficient according to the disturbance noise signal in the cockpit under the current working condition, so as to control the noise cancellation speaker to perform noise cancellation, reduce the amount of noise cancellation calculation and convergence time, and reduce the chip computing power.

[0095] In the description of this specification, any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code including one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present invention belong.

[0096] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0097] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0098] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0099] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0100] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present invention. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

[0101] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, substrates, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example.

[0102] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A noise control method, characterized in that: include: Determine a target filter coefficient in a pre-trained filter database that matches a disturbance noise signal in the cabin under current operating conditions; Noise cancellation control is performed according to the target filter coefficient.

2. The noise control method according to claim 1, characterized in that: The pre-trained filter database includes filter coefficients corresponding to noises of multiple different frequency bands pre-trained offline.

3. The noise control method according to claim 1 or 2, characterized in that: The determining of a target filter coefficient in a pre-trained filter database that matches a disturbance noise signal in the cabin under current working conditions includes: Obtaining an index value of the target filter coefficient according to the disturbance noise signal; The target filter coefficients are determined from the pre-trained filter database according to the index value.

4. The noise control method according to claim 3, characterized in that: The step of obtaining the index value of the target filter coefficient according to the disturbance noise signal comprises: A target noise frequency band matching the disturbance noise signal is determined, and an index value of a filter coefficient in the pre-trained filter database corresponding to the target noise frequency band is obtained as an index value of the target filter coefficient.

5. The noise control method according to claim 4, characterized in that: The determining of a target noise frequency band matching the disturbance noise signal and obtaining an index value of a filter coefficient in the pre-trained filter database corresponding to the target noise frequency band includes: The disturbance noise signal is subjected to frequency band classification by a frequency band classification neural network model to determine the target noise frequency band and output the index value of the target filter coefficient.

6. The noise control method according to claim 5, characterized in that: The frequency band classification neural network model includes a one-dimensional convolutional neural network, which takes the disturbance noise signal as input and takes the index value of the target filter coefficient as output.

7. The noise control method according to claim 6, characterized in that: The one-dimensional convolutional neural network is configured with a residual block.

8. The noise control method according to claim 5, characterized in that: The frequency band classification neural network model includes a two-dimensional convolutional neural network, which takes the disturbance noise signal as input and takes the index value of the target filter coefficient as output.

9. The noise control method according to claim 1 or 2, characterized in that: The performing noise removal control according to the target filter coefficient comprises: The noise cancelling speaker is controlled to play according to the target filter coefficient, so that the sound signal played by the noise cancelling speaker reaches the disturbance noise signal collection location and cancels out the disturbance noise signal.

10. The noise control method according to claim 9, characterized in that: The signal size of the sound signal played by the noise cancellation speaker arriving at the disturbance noise signal collection point is expressed as the product of the external noise reference signal, the target filter coefficient and the sound signal played by the noise cancellation speaker.

11. The noise control method according to claim 10, characterized in that: The noise control method further comprises: The external noise reference signal is determined according to the device speed under the current working condition.

12. The noise control method according to claim 11, characterized in that: The determining the external noise reference signal according to the device speed under the current working condition includes: When the speed of the device is greater than the speed threshold, the external noise reference signal is a wind noise signal collected by a wind noise collection device; Alternatively, when the device speed is less than or equal to the speed threshold, the external noise reference signal is a road noise signal collected by a road noise collection device.

13. The noise control method according to claim 10, characterized in that: The noise control method further comprises: The external noise reference signal is determined according to a correlation coefficient between the external noise signal and the disturbance noise signal.

14. The noise control method according to claim 13, characterized in that: The external noise signal includes a road noise signal collected by a road noise collection device and a wind noise signal collected by a wind noise collection device; Determining the external noise reference signal according to the correlation coefficient between the external noise signal and the disturbance noise signal includes: When the correlation coefficient between the road noise signal and the disturbance noise signal is greater than the correlation coefficient between the wind noise signal and the disturbance noise signal, the external noise reference signal is the road noise signal; Alternatively, when the correlation coefficient between the wind noise signal and the disturbance noise signal is greater than the correlation coefficient between the road noise signal and the disturbance noise signal, the external noise reference signal is the wind noise signal.

15. An electronic device, characterized in that: include: at least one processor; a memory communicatively coupled to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, and when the computer program is executed by the at least one processor, the noise control method according to any one of claims 1 to 14 is implemented.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the noise control method according to any one of claims 1 to 14 is implemented.

17. A movable device, characterized in that: The drivable device comprises the electronic device according to claim 15; Alternatively, the traversable device comprises: A first collecting device is arranged in the cabin at a position corresponding to a human ear, and is used to collect a disturbance noise signal in the cabin; A second collecting device is used to collect external noise signals; A noise cancelling speaker and a controller, wherein the controller is connected to the first collection device, the second collection device and the noise cancelling speaker, and is used to control the noise cancelling speaker according to the noise control method according to any one of claims 1-14.

Images