Reference sensor optimization method for road noise active control system

Abstract

This invention discloses a reference sensor optimization method for an active road noise control system. The method includes the following steps: (1) configuring the vehicle hardware, including deploying sensors, road noise detection microphones, and cancelling speakers, and connecting a multi-channel signal acquisition unit and a real-time road noise controller; (2) determining the driving conditions and the number of preferred road noise references, and establishing a reference optimization database and test set; (3) based on a strong causal iterative reference sensor optimization strategy, selecting the vibration reference combination with the best road noise control performance under A-weighting, and calculating the fixed control filter coefficients by combining the road noise signal collected at the road noise detection microphone; (4) importing the filter coefficients into the road noise controller for real-time noise reduction at the cabin noise reduction point. The method of this invention can be used in an active road noise control system for automobiles based on a feedforward strategy, selecting a reference sensor combination with the best road noise control performance while ensuring the causality of the system.

Description

Technical Field

[0001] This invention belongs to the technical field of active noise control, specifically relating to a reference sensor optimization method for an active road noise control system. Background Technology

[0002] With the rapid development of the automotive industry, NVH (Noise, Vibration, Harshness) performance has become an important indicator for evaluating vehicle comfort. Road noise is one of the main sources of vehicle noise, and it has broadband random noise characteristics that are affected by road conditions. Therefore, Active Road Noise Control (ARNC) technology is an important direction for the automotive NVH industry. In actual ARNC systems, accelerometers are placed under the chassis as reference sensors to pick up vibration information as reference signals. The use of feedforward multi-channel FxLMS algorithm for noise reduction at the human ear in the cabin has been verified as a reliable technical solution (Sutton TJ, Elliott SJ, McDonald AM, et al. Active control of roadnoise inside vehicles[J]. Noise Control Engineering Journal,1994,42(4):137-147.). More reference signals can maintain better coherence with the road noise signal at the human ear, thus achieving higher noise reduction. However, this will place a heavy burden on the system's real-time performance and computational load. Therefore, using a reasonable number of reference signals and selecting the optimal combination of reference signals while ensuring noise reduction performance are key issues for ARNC systems that adopt a feedforward strategy.

[0003] Currently, the Multiple Coherence (MC) method is widely used to calculate the multiple coherence coefficients of the reference signal and the road noise signal at each frequency point to estimate the maximum noise reduction. The reference signal is then optimized based on the coherence of the two signals (OhS H, Kim H, Park Y. Active control of road booming noise in automotive interiors[J].The Journal of the Acoustical Society of America,2002,111(1):180-188.). However, in actual road noise control, the causality of the system has a significant impact on the noise reduction. Due to physical limitations, highly coherent vehicle vibration signals may be generated by passive excitation sources, which lag behind the road noise signal and thus exhibit non-causal characteristics. It is impossible to track and predict the road noise signal. In other words, strong coherence cannot guarantee causality, and the reference combination with the best coherence cannot achieve the physically achievable maximum noise reduction. Therefore, how to select the optimal reference signal combination with the best control performance while ensuring the causality of the system has become the primary problem for reference optimization and noise control in ARNC systems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a reference sensor optimization method for an active noise control system under a strong causal iterative strategy. The strong causal strategy is based on the Wiener filtering principle, estimating the maximum noise reduction while considering the causality of the actual physical system. The iterative strategy obtains local optimal solutions by progressively expanding the autocorrelation matrix of the filtered reference signal and its cross-correlation matrix with the desired signal; the overall optimal solution is obtained when the optimal reference combination is reached.

[0005] The technical solution adopted in this invention is as follows:

[0006] A method for optimizing the reference sensor in an active road noise control system includes the following steps:

[0007] Step 1: Configure the vehicle hardware, including deploying sensors, road noise detection microphones and cancelling speakers, and connecting a multi-channel signal acquisition unit and a real-time road noise controller.

[0008] Step 2: Determine the vehicle driving conditions and the optimal number of road noise references, and establish a reference optimization database and test set;

[0009] Step 3: Based on the strong causal iterative reference optimization strategy, select the vibration reference combination with the best road noise control performance under A weighting, and calculate the fixed control filter coefficients by combining the road noise signal collected at the road noise detection microphone.

[0010] Step 4: Import the filter coefficients into the real-time road noise controller to perform real-time noise reduction in the noise reduction area of ​​the vehicle cabin.

[0011] Furthermore, step 2 is specifically implemented as follows: determining the vehicle driving conditions and the optimal number of road noise references; during vehicle operation, a multi-channel signal acquisition unit acquires P reference signals x. i And the desired signal d at the road noise detection microphone j Establish a reference optimization database and a reference combination generalization test set; use the reference optimization database to optimize the Q-path reference signal combination from the P-path reference signals; use the test set to perform generalization tests on the optimized combination; where i = 1, 2, ..., P, j = 1, 2.

[0012] Compared with the prior art, the beneficial effects of the present invention are: in the reference selection of automotive ARNC, it has the ability to select the reference signal combination with the maximum noise reduction under the premise of ensuring system causality, thereby achieving better road noise control effect. Attached Figure Description

[0013] Figure 1 This is an overall structural block diagram of the method of the present invention.

[0014] Figure 2 The following are hardware configuration diagrams for the method of the present invention: (a) is a schematic diagram of the placement of the vehicle chassis acceleration vibration sensor, and (b) is a schematic diagram of the installation of the active noise reduction headrest in the vehicle cabin.

[0015] Figure 3 This is a flowchart of the process using a strong causal iterative reference optimization strategy.

[0016] Figure 4 These are the time-domain impulse response and amplitude-frequency response curves of the secondary path of the active noise-canceling headrest in the embodiment. (a) and (b) are the time-domain impulse response and amplitude-frequency response curves of the secondary path from the left speaker to the left and right ear noise detection microphones; (c) and (d) are the time-domain impulse response and amplitude-frequency response curves of the secondary path from the right speaker to the left and right ear noise detection microphones.

[0017] Figure 5 This is a diagram of the fixed control filter weight coefficients calculated in the embodiment.

[0018] Figure 6 These are the time-frequency domain noise reduction curves at the vehicle cabin road noise detection microphone in the embodiment. (a) and (b) are the time-frequency domain noise reduction curves at the left ear, and (c) and (d) are the time-frequency domain noise reduction curves at the right ear. Detailed Implementation

[0019] The reference sensor optimization method for the active path noise control system under the strong causal iterative strategy of this invention mainly includes the following parts:

[0020] 1. Hardware Configuration

[0021] 1) Installation of acceleration sensor in automobile chassis

[0022] Feedforward ARNC systems require road noise vibration source information as reference signals. Therefore, accelerometers (single-axis or multi-axis) need to be deployed on the vehicle chassis to pick up vehicle vibration fluctuation information. The reference signal used in ARNC should have strong coherence with the road noise signal at the road noise detection microphone in the cabin, thereby achieving a large noise reduction. The paper (Park YS, Cho MH, Oh CS, et al. Coherence-based sensor set expansion for optimal sensor placement in active road noise control[J]. Mechanical Systems and Signal Processing, 2022, 169: 108788.) places accelerometers in locations such as the subframe (zone 1), steering shaft (zone 2), shock absorber (zone 3), and trailing arm (zone 4) to obtain reference signals with strong coherence, such as... Figure 2 As shown in (a).

[0023] 2) Active noise-canceling headrests in the vehicle cabin

[0024] like Figure 2 As shown in (b), the active noise-canceling headrest includes two road noise detection microphones (M1 and M2) simulating the quiet zone of the human ear, and canceling speakers (S1 and S2). Each of the two road noise detection microphones is placed beside the user's ears, providing reliable audio hardware support for the ARNC system. On the one hand, the low-frequency response of the speakers should be relatively flat to ensure good low-frequency response in the secondary path between the speakers and microphones; on the other hand, the headrest should be as comfortable and aesthetically pleasing as possible to provide passengers with a better riding experience. Specific functions are as follows:

[0025] First, during the secondary path modeling process, the controller generates white noise as a reference signal to drive the secondary source (cancellation speaker) to emit sound. This sound is then collected by a microphone at the human ear and input into the controller as the desired signal. The controller calculates and matches the secondary path from the secondary source to the microphone. Once the secondary path modeling is complete, it is exported for use in the selection of reference combinations and the calculation of fixed control filter coefficients.

[0026] Second, in the process of selecting the reference combination, a road noise detection microphone is used to pick up road noise signals from both ears, and the selection operation is carried out in combination with the road noise signals collected by the multi-channel acquisition device.

[0027] Third, during the real-time road noise control process, the noise at the ear is canceled out by the sound emitted by the cancelling speaker, so as to realize the real-time noise reduction function of the noise reduction area in the cabin.

[0028] 2. Calculation of reference selection and fixed control filter coefficients

[0029] 1) Noise Reduction Statistical Strategy and Principle

[0030] The principle of the widely used multicoherent method for noise reduction estimation is as follows.

[0031] The multicoherence coefficients of the reference signal and the road noise signal at each frequency point are calculated. Incoherent output can be represented as:

[0032]

[0033] In equation (1), S dd (f) is the power spectrum of the road noise signal, S nn (f) represents the residual noise spectrum at the road noise detection microphone that is uncorrelated with the reference signal. Therefore, the maximum noise reduction of the road noise signal by the reference signal can be estimated using the incoherent output, i.e.:

[0034]

[0035] The maximum noise reduction is estimated by calculating the multicoherence coefficients of the reference signal and the road noise signal at each frequency point through the above process. The reference selection is based on the coherence between the reference signal and the road noise signal, but the causality of the actual physical system is not considered.

[0036] The statistical principle of noise reduction for the strongly correlated iterative reference optimization strategy proposed in this invention is as follows.

[0037] Assuming the number of accelerometers, cancelling loudspeakers, and road noise detection microphones in the ARNC system are K, M, and L respectively, then the control filter weight coefficients have a total of M×K sets, and the secondary paths have a total of L×M sets. lm =[s lm (0)s lm (1)…s lm (J-1)] T This represents the secondary path from the m-th loudspeaker to the l-th path noise detection microphone, modeled using a J-order FIR filter; w mk =[w mk (0)w mk (1)...w mk (I-1)] T Let I represent the control filter weight vector from the k-th reference sensor to the m-th loudspeaker, with a total order of I.

[0038] xk (n) represents the k-th reference signal at time n, then the output signal of the m-th loudspeaker is:

[0039]

[0040] At the road noise detection microphone, the l-th error signal e l (n) is:

[0041]

[0042] In equation (4), d l (n) represents the path noise signal at the l-th microphone. Reference signal x k (n) The filtered reference signal after passing through the ml-th secondary path can be expressed as:

[0043]

[0044] Substituting equations (3) and (5) into equation (4), we get

[0045]

[0046] Therefore, the error signal vector e(n) can be expressed as:

[0047]

[0048] In equation (7), the road noise signal vector d(n) is:

[0049] d(n) = [d1(n) d2(n) … d L (n)] T (8)

[0050] The filter weight vector W(i) for a single tap is written as:

[0051] W(i) = [w 11 (i) w 12 (i) … w 1K (i) w 21 (i) … w MK (i)] T (9)

[0052] The filter reference signal matrix at time n is:

[0053]

[0054] Furthermore, the weight vector W(i) of the individual tap combinations is combined into the form of length MKI, that is:

[0055] W = [W T (0) WT (1) … W T (I-1)] T (11)

[0056] The corresponding composite matrix of the filtered reference signal can be written as:

[0057]

[0058] At this point, equation (7) can be rewritten as:

[0059]

[0060] Let the objective function be the mean squared error, then:

[0061]

[0062] In equation (14), E(·) represents the time average of the independent variable.

[0063] It is generally assumed that the road noise signal is stationary, and J is a quadratic function of the filter weight vector W. Setting the gradient of the mean square error to zero, the Wiener solution for the weights is obtained as follows:

[0064] W opt =-R -1 p (15)

[0065] In equation (15),

[0066] W opt Substituting into equation (13), based on the road noise signal and the filtered reference signal, the error signal obtained from the Wiener solution can be obtained:

[0067]

[0068] At this point, the average noise reduction can be expressed as:

[0069]

[0070] 2) Reference Combination Optimization and Generalization

[0071] The optimal road noise reference combination is considered as the overall optimal solution. The process of solving for the noise reduction corresponding to different reference combinations is divided into several sub-problems. Solving each sub-problem yields a local optimal solution (i.e., the reference combination corresponding to the maximum noise reduction). Combining the local optimal solutions of the sub-problems yields the overall optimal solution. Following this idea, firstly, the reference signals with the maximum noise reduction under causal conditions are selected; secondly, the autocorrelation matrix of the filtered reference signals and its cross-correlation matrix with the road noise signal are progressively expanded until the reference signal combination with the maximum noise reduction for the desired channel is selected. The experimental procedure is as follows... Figure 3 As shown.

[0072] To ensure that the selected road noise reference combination has strong universality, reference signal combinations selected from different datasets in the database were substituted into the test set for Wiener noise reduction test, and the combination with the largest noise reduction was used as the optimal reference combination.

[0073] 3. The fixed control filter coefficients obtained by Wiener filtering using the preferred reference combination are imported into the controller. The vibration signal of the car chassis collected by the acceleration sensor is used as a reference signal and sent to the controller. After being output by the control filter, the cancelling speaker of the active noise-reducing headrest is driven to produce sound, so as to realize the function of real-time noise control in the noise reduction area of ​​the cabin.

[0074] Example

[0075] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0076] 1. Placement of accelerometers on the vehicle chassis

[0077] The vehicle model used in this experiment is the Changan UNI-K. A total of 20 3-axis accelerometers were deployed to pick up P=60 reference signals. The goal was to select the optimal combination of Q=16 reference signals for noise control in the noise reduction area inside the vehicle cabin.

[0078] 2. Calculation of reference selection and fixed control filter coefficients

[0079] 1) Obtaining secondary paths

[0080] The active noise-canceling headrest designed in this invention consists of a left ear speaker (Speaker1, S1) and a right ear speaker (Speaker2, S2), a left ear microphone (Microphone1, M1) and a right ear microphone (Microphone2, M2) simulating the noise reduction point of the human ear, as follows. Figure 2 As shown in (b). The sound emitted by S1 is collected by M1 and M2 to obtain secondary paths S11 and S12, and the sound emitted by S2 is collected by M1 and M2 to obtain secondary paths S21 and S22. To ensure the secondary path modeling is as accurate as possible, a quiet road section without noise interference should be selected during measurement, and the car engine should be turned off. To simulate real-world driving conditions, a tester should be stationed in the passenger seat. The 512th-order time-domain impulse response and amplitude-frequency response of secondary paths S11 and S12 are shown below. Figure 4 As shown in (a) and (b); the 512th-order time-domain impulse response and amplitude-frequency response of secondary paths S21 and S22 are as follows. Figure 4 As shown in (c) and (d).

[0081] 2) Strong causal iterative reference combination optimization

[0082] First, a reference optimization database and test set are established. In this embodiment, the vehicle maintains a steady-state driving speed of 60 km / h. Six sets of 30-second reference signal libraries and road noise signal libraries are recorded, along with one set of 120-second reference signal set and road noise signal set. Each set of data contains 60 reference signals x1(n) to x... 60 (n) and two noise signals d1(n)~d2(n) are used for Wiener filtering to select the optimal combination of reference signals. After control, the error signals at the left and right ears are e1(n)~e2(n).

[0083] Secondly, the filter order I in the Wiener filter is set to 512. Six reference combinations are selected from eight datasets. The six reference combinations are then substituted into the test set for Wiener filtering and the average noise reduction at the left and right ears is calculated. The calculation method is as shown in equation (18). The group with the largest noise reduction is selected as the optimal reference combination for real-time noise reduction at the vehicle cabin road noise detection microphone.

[0084]

[0085] To demonstrate the performance improvement of this invention compared to existing methods, this embodiment compares and verifies the noise reduction achieved by using Wiener filtering and multiple coherence methods for reference selection. The experimental results are shown in Tables 1 and 2. It was found that the reference signal combinations selected using Wiener filtering consistently outperformed those selected using multiple coherence methods in terms of noise reduction for the test set. Based on the comparison in Table 1, group 1 was selected as the final preferred path noise reference combination, with the selected reference numbers being 50, 40, 22, 59, 5, 20, 39, 24, 55, 37, 51, 4, 1, 45, 36, and 44.

[0086] Table 1. Optimal Results of Wiener Filtering Method

[0087]

[0088] Table 2. Optimization Results of the Multicoherence Method

[0089]

[0090] 3. Real-world vehicle noise reduction verification

[0091] 1) Obtaining the coefficients of the fixed control filter

[0092] Wiener filtering calculations were performed using the reference signal, path noise signal, and secondary path derived from the controller, with a data recording length of 30 seconds. For a 16*2*2 ANC system, when the control filter length is 512, a total of 32*512 filter coefficients were obtained, such as... Figure 5As shown. In order to avoid the influence of low-frequency components in the signal (below 50Hz, which can be regarded as the DC component in the signal) on the control filter, all signals are passed through a 50Hz high-pass filter during the calculation.

[0093] 2) Road noise control performance test based on A-weighting

[0094] Under the same road conditions and vehicle speed, 30 seconds of data were recorded at the error microphone with ARNC off and ARNC on, respectively, as the desired signal and the processed error signal. The sound pressure level difference between the two under A-weighted statistics was compared, and the average noise reduction within 30 seconds for e1 (left ear) was 6.06 dBA, and the average noise reduction within 30 seconds for e2 (right ear) was 5.01 dBA. The time-frequency domain noise reduction results are as follows: Figure 6 As shown.

[0095] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing reference sensors in an active road noise control system, characterized in that, The method includes the following steps: Step 1: Configure the vehicle hardware, including deploying sensors, road noise detection microphones and cancelling speakers, and connecting a multi-channel signal acquisition unit and a real-time road noise controller. Step 2: Determine the vehicle driving conditions and the optimal number of road noise references, and establish a reference optimization database and test set; Step 3, the Wiener solution expression obtained based on the strong causal iterative reference optimization strategy is: in, for The autocorrelation matrix, where n is the time index. This indicates that the independent variable is averaged over time. This is the composite matrix of the filter reference signal; for and The cross-correlation matrix, Let the road noise signal vector be denoted as ; then the error signal vector is represented as: ， The vibration reference combination with the optimal road noise control performance under A-weighting is selected, and the coefficients of the fixed control filter are calculated based on the road noise signal collected at the road noise detection microphone; specifically: First, the coefficients of the fixed control filter under a single reference are calculated according to the Wiener solution expression to obtain the A-weighted noise reduction amount at this time, and the reference signal with the maximum noise reduction amount under causal conditions is selected. Second, the autocorrelation matrix of the filtered reference signal and its cross-correlation matrix with the path noise signal are gradually expanded until the reference signal combination with the maximum noise reduction amount of the desired channel is selected. In order to ensure that the selected reference signal combination has strong universality, the reference signal combinations selected from different datasets in the reference optimization database are substituted into the test set for Wiener noise reduction test, and the combination with the maximum noise reduction amount is used as the optimal reference signal combination. Step 4: Import the filter coefficients into the real-time road noise controller to perform real-time noise reduction in the noise reduction area of ​​the vehicle cabin.

2. The reference sensor optimization method for an active road noise control system as described in claim 1, characterized in that, In step 1, an acceleration sensor is placed on the vehicle chassis to pick up road vibration information and collect multiple reference signals.

3. The reference sensor optimization method for an active road noise control system as described in claim 1, characterized in that, In step 1, an active noise-canceling headrest is installed in the vehicle cabin. The active noise-canceling headrest includes two road noise detection microphones that simulate the quiet zone of the human ear and a canceling speaker. The two road noise detection microphones are each placed next to the person's ears.

4. The reference sensor optimization method for an active road noise control system as described in claim 1, characterized in that, The specific implementation method of step 2 is as follows: determine the vehicle driving conditions and the optimal number of road noise references, and collect P-channel reference signals by a multi-channel signal acquisition unit during vehicle operation. And the desired signal at the road noise detection microphone A reference optimization database and a reference combination generalization test set are established; the reference optimization database is used to optimize the Q-path reference signal combination from the P-path reference signals; and the test set is used to perform generalization tests on the optimized combination; among these, , .

5. The reference sensor optimization method for an active road noise control system as described in claim 1, characterized in that, In step 4, the filter coefficients are imported into the real-time road noise controller. The vibration signal of the vehicle chassis collected by the sensor is sent to the real-time road noise controller as a reference signal. After being output by the control filter, the cancellation speaker is driven to emit sound, thereby realizing the function of real-time noise control in the noise reduction area of ​​the vehicle cabin.

Images