A window noise reduction system and method based on a multi-error ANC algorithm under the Minimax criterion

Abstract

This invention discloses a window noise reduction system based on a multi-error ANC algorithm under the Minimax criterion, comprising: a window body and a multi-channel active noise control system. The multi-channel active noise control system includes: multiple reference microphones located outside the wall where the window body is located, evenly distributed on both sides of the window of the window body, for collecting raw outdoor noise signals; multiple secondary speakers located indoors and evenly distributed on both sides of the window of the window body, for controlled output of anti-noise signals for noise reduction from indoors; multiple error microphones located indoors and fixed on error microphone holders; multiple error microphones evenly distributed in front of multiple secondary speakers for collecting indoor noise signals; and a controller that implements the multi-error ANC algorithm based on the Minimax criterion to control indoor noise. The multi-error ANC algorithm based on the Minimax criterion uses a normalization method based on the Minimax criterion combined with the tanh function as the control algorithm. A corresponding noise reduction method is also disclosed.

Description

Technical Field

[0001] This invention relates to the field of active noise reduction control technology for indoor noise control, and in particular to a window noise reduction system and method based on a multi-error ANC algorithm under the Minimax criterion. Background Technology

[0002] With societal progress, more and more people are flocking to cities. Along with this population growth, urban noise problems are becoming increasingly severe, making it increasingly difficult to sit quietly indoors to study or work. Furthermore, prolonged exposure to noise increases health risks. To ensure that environmental noise levels meet health standards, governments worldwide are continuously exploring noise control measures to protect public health and are attempting to formally incorporate the protection of the sound environment into legal safeguards, implementing regulations to restore quiet to the people and protect a harmonious and peaceful living environment.

[0003] Most common noise control methods are passive noise control (PNC), which primarily works by attenuating sound energy through the interaction of noise waves with acoustic materials or structures, thereby reducing noise levels. While closing windows can block some noise, it also reduces indoor ventilation, impacting indoor air quality. Furthermore, it is ineffective attenuating low-frequency noise. With the ever-increasing urban population, effectively reducing indoor noise while ensuring ventilation is crucial for protecting the health of the growing urban population. Active noise control (ANC) technology effectively compensates for the shortcomings of PNC, demonstrating excellent results in suppressing low-frequency noise and representing a new research area in noise reduction. Currently, active noise control (ANC) has proven to be a practical noise reduction solution. Its mechanism is based on the principle of destructive interference in acoustic theory. It generates a sound wave signal with the same frequency and amplitude but opposite phase as the original noise signal, and the two signals undergo destructive interference to eliminate the noise. ANC effectively attenuates low-frequency noise and has great potential for noise reduction over large spaces. The system has a simple structure, making it easy to design, install, and control. In recent years, ANC technology has achieved noise reduction in relatively small spaces, such as in noise-canceling headphones and some active noise cancellation systems in cars. However, its application in indoor spaces is virtually nonexistent. For large-scale indoor noise reduction, to achieve better results, the active noise control system needs to expand the number of channels. As the number of channels increases, the computational load of the multi-channel control algorithm also increases significantly, impacting the controller's resource consumption and the system's real-time performance.

[0004] Currently, the implementation of ANC technology systems can be divided into two main categories: Digital Signal Processors (DSPs) and Field-Programmable Gate Arrays (FPGAs). In the early days, DSPs could meet the computational requirements for ANC technology implementation. However, as the complexity of systems continues to increase, FPGAs have gained widespread attention due to their advantages in parallel processing and strong computing power. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention provides the following technical solution: a window noise reduction system and method based on a multi-error ANC algorithm under the Minimax criterion. In actual operation, an indoor noise threshold is set within the system. When the noise reduction system is active, multiple reference microphones located outdoors collect outdoor noise signals, and multiple error microphones located indoors collect indoor noise signals. The noise value is calculated and compared with the threshold. When the indoor noise value exceeds the threshold, the active noise control process begins. Multiple secondary sound sources are controlled to emit sound, and related secondary paths are modeled. After modeling is completed, the active noise control algorithm is executed to update the controller parameters in real time, thereby generating a secondary signal with certain characteristics for noise reduction. The convergence speed and steady-state performance of the algorithm are improved by introducing the Minimax criterion and a normalization method improved by combining the tanh function. Furthermore, FPGA technology is used to enhance parallel processing and computational capabilities.

[0006] This invention provides a window noise reduction system based on a multi-error ANC algorithm under the Minimax criterion, comprising: a window body (1) and a multi-channel active noise control system, wherein:

[0007] The multi-channel active noise control system includes:

[0008] Multiple reference microphones (4) are located on the outside of the wall where the window body (1) is located, and are evenly distributed on both sides of the window of the window body (1) to collect the original noise signal from the outside.

[0009] Multiple secondary speakers (7) are located indoors and evenly distributed on both sides of the window body (1) for controlled output of anti-noise signals to reduce noise from indoors;

[0010] Multiple error microphones (5) are located indoors and fixed on an error microphone mounting bracket (8); the multiple error microphones (5) are evenly distributed in front of the multiple secondary speakers (7) for collecting indoor noise signals; and

[0011] The controller controls indoor noise by implementing a multi-error ANC algorithm based on the Minimax criterion; the multi-error ANC algorithm based on the Minimax criterion uses normalization based on the Minimax criterion combined with the tanh function as the control algorithm.

[0012] Preferably, the controller is an FPGA processor. An STM32 microcontroller is used to configure the I2C communication protocol registers of the ADC and DAC chips, setting the ADC and DAC chips as slaves and the data transmission format as I2S. The FPGA acts as the master, generating input and output I2S clock signals and applying timing constraints. A reference microphone collects signals from the noise source, and an error microphone array collects error signals at the noise cancellation point. These signals are converted into digital signals by the ADC chip and enter the FPGA processor via the I2S audio interface. The FPGA processor processes the digital signals to generate output signals with specific characteristics, which are then converted into analog signals by the DAC. After power amplification, these signals drive the secondary sound source to emit sound to cancel out the noise at the noise cancellation point. A ROM core stores the coefficients of the fixed filter, i.e., the estimated transfer function of the secondary path. A dual-port RAM core is used to buffer the input data of the fixed filter.

[0013] Preferably, it also includes a remote controller (2) for controlling the opening or closing of the multi-channel active noise control system.

[0014] Preferably, it also includes a power indicator light (3), which is used to display the working status of the multi-channel active noise control system; when the multi-channel active noise control system is in the open state, the power indicator light (3) is always on.

[0015] Preferably, it also includes a display screen (6) for displaying indoor and outdoor noise levels respectively; acquiring multiple outdoor original noise signal values ​​through multiple reference microphones (4), calculating the average value of the multiple outdoor original noise signal values ​​as the outdoor noise level and displaying it on the display screen (6); acquiring multiple indoor noise signal values ​​through multiple error microphones, calculating the average value of the multiple indoor noise signal values ​​as the indoor noise level and displaying it on the display screen (6).

[0016] A second aspect of the present invention is to provide a windowing denoising method based on a multi-error ANC algorithm under the Minimax criterion, comprising:

[0017] S1, detects the ambient noise level;

[0018] S2, Model the secondary path of the secondary speaker (7) and the error microphone (5) to obtain the secondary path model parameters;

[0019] S3, based on the secondary path model parameters, performs noise reduction through the controller, thereby controlling indoor noise.

[0020] Preferably, S1 includes:

[0021] S11, the first dataset is obtained by collecting the original outdoor noise signal through multiple reference microphones (4), and the first dataset is sent to the controller to calculate the average outdoor noise level and then displayed as the outdoor noise level. The outdoor noise level will be used as a noise reduction reference.

[0022] S12: The second dataset is obtained by collecting indoor noise signals through multiple error microphones (5). The second dataset is sent to the controller to calculate the average indoor noise level and then displayed as the indoor noise level. The indoor noise level will be used as the condition for the noise reduction system to be turned on.

[0023] S13: Set a noise reduction start threshold in the active noise control system. When the indoor noise level is detected to be greater than the noise reduction start threshold, noise reduction is started.

[0024] Preferably, S2 includes:

[0025] Step 21: Control the secondary speaker (7) to emit sound, and use the LMS algorithm to model the secondary path between the secondary speaker (7) and the error microphone (5); calculate the secondary path error value;

[0026] Step 22: Repeat step S21 to obtain multiple channel error values ​​through multiple iterations. Take the channel error value with the smallest value as the secondary path model parameter and store it in the controller.

[0027] Preferably, S3 includes:

[0028] S31: The outdoor raw noise signal and indoor noise signal, which are collected in real time through multiple reference microphones and multiple error microphones, are respectively used as the input signal and error signal of the controller;

[0029] S32: The noise reduction algorithm is executed via FPGA based on the secondary path model parameters;

[0030] S33: The output signal is sent by driving multiple secondary speakers (7) and superimposed on the sound wave at the error microphone to achieve noise reduction.

[0031] Preferably, the noise reduction algorithm in S32 includes:

[0032] The adaptive filter coefficients are updated according to formulas (12) and (13) to generate an output signal with the same amplitude but opposite phase as the noise signal; where:

[0033] w j (n+1)=w j (n)-μ(n)e m (n)x′ mj (n) (12);

[0035]

[0036] In equation (12), m represents the error microphone label of the maximum error signal; j represents the secondary sound source label; x′ mj (n) represents the estimated secondary path transfer function of the reference signal after passing through the secondary sound source corresponding to the maximum error microphone. The filtered output after; e m (n) represents the signal vector with the largest error signal amplitude among the K error microphones; w j (n+1) and w j (n) represents the weight coefficient matrix of the transverse FIR adaptive filter; in equation (13), α is used to adjust the problem of large step size caused by the small inner product value of the input signal; β is used to adjust the value of the tanh function to ensure that it is between 0 and 1; x(n) represents the vector form of the output signal of the reference microphone at time n.

[0037] The noise reduction system and method provided by this invention have the following beneficial effects:

[0038] 1) This invention implements a window noise reduction system based on an FPGA platform, applying active noise reduction technology to achieve large-scale noise reduction indoors. The placement of secondary sound sources is optimized, employing an edge distribution method to ensure that noise reduction is achieved when windows are open without affecting indoor ventilation.

[0039] 2) The active noise control system proposed in this invention includes a threshold setting and display function, which can automatically detect the noise level through outdoor and indoor microphones, control the noise control switch according to whether the noise change exceeds the threshold, maintain the indoor noise at a relatively stable level, provide a comfortable indoor environment, and intuitively display the noise change through the display screen.

[0040] 3) This invention proposes an algorithm based on the Minimax criterion suitable for multi-error ANC systems as the control algorithm for multi-channel systems. The algorithm updates the filter using only the signal from one error microphone during each filter iteration, significantly reducing computational load compared to the traditional multi-error FxLMS algorithm, thus facilitating controller implementation. It can maximize the number of channels in the system while ensuring the controller's computational capabilities meet requirements, thereby generating a wider noise reduction range. The reduced computational load provides a reference for adding channels to active noise control systems for more complex application scenarios.

[0041] 4) To address the slow convergence speed of the multi-error ANC algorithm using the Minimax criterion, a normalization method using a transformable tanh function is employed to adjust the step size update during the adaptive process. This ensures that the algorithm obtains a larger step size in the early stages of convergence to accelerate the convergence speed, while maintaining a smaller step size in the later stages to reduce steady-state error. Results show that both the convergence speed and steady-state error of the algorithm are further improved.

[0042] 5) In the control algorithm, this invention employs a low-rate sampling data throughput and high-rate filtering calculation design method for the software implementation of the involved filters. Specifically, the data sampling clock frequency is much lower than the FPGA's operating clock frequency. Within a single sampling cycle, a multiplier and RAM core are multiplexed, a dual-port RAM is configured to replace registers for storing filter input data, and a single-port ROM is configured for storing filter tap coefficients. This achieves the filtering process of high-order coefficients with lower resource consumption, greatly saving on-chip resources and facilitating subsequent channel expansion for larger-scale applications. For model-based algorithms, in terms of program design, only one multiplier is used during filter implementation, and data is cached and stored through storage units, significantly reducing FPGA resource consumption and enabling the FPGA algorithm implementation to meet more complex application scenarios.

[0043] 6) The filter is implemented based on a serial filter using dual-port RAM. The entire filter module consists of a read / write control module, dual-port RAM, coefficient ROM, and multiplier. The dual-port RAM is used for buffering input data, and the coefficient ROM is used for storing filter coefficients. The read / write control module generates read / write address counts to control the filtering process. This method can achieve the highest possible filter order, enabling the noise reduction system to calculate higher-order filters for noise control output and achieve better noise reduction effect. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of an active noise reduction window model provided according to a preferred embodiment of the present invention.

[0045] Figure 2 This is a side view of an active noise window model provided according to a preferred embodiment of the present invention.

[0046] Figure 3 A block diagram of a normalized multi-error ANC algorithm based on the Minimax criterion combined with the tanh function, provided according to a preferred embodiment of the present invention.

[0047] Figure 4 This is a schematic diagram of an active noise control process according to a preferred embodiment of the present invention.

[0048] Figure 5 This is a schematic diagram of a control system according to a preferred embodiment of the present invention.

[0049] Figure 6 This is a schematic diagram of the FPGA program structure provided according to a preferred embodiment of the present invention.

[0050] Figure 7(a) shows the noise reduction effect diagram and noise reduction effect curve at the error microphones numbered 1-4 according to the preferred embodiment of the present invention, based on two algorithms and a window model.

[0051] Figure 7(b) shows the noise reduction effect diagram and noise reduction effect curve at the error microphones numbered 5-8 according to the preferred embodiment of the present invention, based on two algorithms and a window model. Detailed Implementation

[0052] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0053] The method provided by this invention can be implemented in a terminal environment that may include one or more of the following components: a processor, a memory, and a display screen. The memory stores at least one instruction, which is loaded and executed by the processor to implement the method described in the following embodiments.

[0054] A processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the terminal, and performs various functions and processes data by running or executing instructions, programs, code sets or instruction sets stored in memory, and by calling data stored in memory.

[0055] Memory can include random access memory (RAM) or read-only memory (ROM). Memory can be used to store instructions, programs, code, code sets, or instructions.

[0056] The display screen is used to show the user interface of each application.

[0057] In addition, those skilled in the art will understand that the structure of the terminal described above does not constitute a limitation on the terminal. The terminal may include more or fewer components, or combine certain components, or have different component arrangements. For example, the terminal may also include radio frequency circuits, input units, sensors, audio circuits, power supplies, and other components, which will not be described in detail here.

[0058] A window noise reduction system and method based on a multi-error ANC algorithm under the Minimax criterion is disclosed. In practical operation, an indoor noise threshold is set within the system. When the noise reduction system is active, multiple reference microphones located outdoors collect outdoor noise signals, while multiple error microphones located indoors collect indoor noise signals. The noise value is calculated and compared with the threshold. When the indoor noise value exceeds the threshold, the active noise control process is initiated. Multiple secondary sound sources are controlled to model related secondary paths. After modeling, the active noise control algorithm is executed to update the controller parameters in real time, thereby generating a secondary signal with specific characteristics for noise reduction. The computational complexity of the algorithm is reduced by introducing the Minimax criterion and a normalization method improved by combining the tanh function. Furthermore, FPGA technology is employed to enhance parallel processing and computational capabilities.

[0059] A multi-channel active noise cancellation (ANC) system was implemented on an FPGA platform and applied to an active noise cancellation window. The layout of related electroacoustic components was optimized, and secondary sound sources were distributed at the edges to provide a comfortable and quiet indoor environment while ensuring ventilation. To reduce the computational load of the multi-channel ANC system, the Minimax criterion combined with tanh function normalization was used as the control algorithm to process the error signals of the multi-error microphones. In each iteration, only the signal from the microphone with the largest amplitude was used for iteration. Compared with the standard multi-error Fxlms algorithm, the computational load was significantly reduced while meeting the convergence speed and steady-state error requirements of practical system implementation. Key steps in the program design based on the FPGA platform were presented, and the filter design method in the program was optimized to reduce resource consumption, providing algorithmic and program design references for the application of larger-scale multi-channel active noise control systems.

[0060] See Figure 1-2 This invention provides a window denoising system based on a multi-error ANC algorithm under the Minimax criterion, comprising:

[0061] Window body 1 and multi-channel active noise control system, wherein:

[0062] The multi-channel active noise control system includes:

[0063] Multiple reference microphones 4 are located outside the wall where the window body 1 is located and are evenly distributed on both sides of the window of the window body 1 for collecting the original noise signals outdoors; in this embodiment, the power lines and data lines supporting the multiple reference microphones 4 are placed in a junction box;

[0064] Multiple secondary speakers 7 are located indoors and are evenly distributed on both sides of the window of the window body 1 for controlled output of anti-noise signals to reduce noise indoors; in this embodiment, the power lines and data lines supporting the multiple secondary speakers 7 are placed in a junction box; by optimizing the layout of the secondary speakers 7 and adopting an edge layout method, it is ensured that the active noise reduction window can achieve noise reduction when the window is open and when the window is closed, and can not affect the ventilation conditions in the room when performing the noise reduction function, so that the noise reduction window can be applicable to the requirements under various climate conditions and meet the usage requirements in different scenarios;

[0065] Multiple error microphones 5 are located indoors and are fixed on the error microphone fixing bracket 8; in this embodiment, a thin iron bracket is used as the error microphone fixing bracket 8; the multiple error microphones 5 are evenly distributed directly in front of the multiple secondary speakers 7 for collecting the noise signals indoors; in this embodiment, the power lines and data lines supporting the multiple error microphones 5 are placed in a junction box; and

[0066] A controller, the controller performs noise control on the indoor environment by implementing a multi-error ANC algorithm based on the Minimax criterion; the multi-error ANC algorithm based on the Minimax criterion uses the normalization combining the minimax criterion and the tanh function as the control algorithm.

[0067] As a preferred implementation manner, it further includes: a remote controller 2 and a power indicator light 3, the remote controller 2 is used to control the opening or closing of the multi-channel active noise control system, and the power indicator light 3 is used to display the working state of the multi-channel active noise control system; when the multi-channel active noise control system is in the open state, the power indicator light 3 will be on constantly.

[0068] As a preferred implementation manner, it further includes: a display screen 6 for respectively displaying the indoor and outdoor noise levels. Among them, indoor and outdoor environmental noise detection is added to the active noise control system and is displayed in real time through the display screen 6. Multiple original noise signal values outdoors are collected through multiple reference microphones 4, and the average value of the multiple original noise signal values outdoors is obtained as the outdoor noise level and is displayed through the display screen 6; multiple indoor noise signal values are collected through multiple error microphones, and the average value of the multiple indoor noise signal values is obtained as the indoor noise level and is displayed through the display screen 6.

[0069] See Figure 3-6The second aspect of the present invention is to provide a windowing denoising method based on a multi-error ANC algorithm under the Minimax criterion, comprising:

[0070] S1, detecting ambient noise levels, including:

[0071] S11, the first dataset is obtained by collecting the original outdoor noise signal through multiple reference microphones 4, and the first dataset is sent to the controller to calculate the average outdoor noise level and then displayed as the outdoor noise level. The outdoor noise level will be used as a noise reduction reference.

[0072] S12: The second dataset is obtained by collecting indoor noise signals through multiple error microphones 5. The second dataset is sent to the controller to calculate the average indoor noise level and then displayed as the indoor noise level. The indoor noise level will be used as the condition for the noise reduction system to be turned on.

[0073] S13: A noise reduction start threshold is set within the active noise control system. When the indoor noise level is detected to be greater than the noise reduction start threshold, noise reduction is initiated. In this embodiment, the condition for noise reduction is that noise reduction begins when the indoor noise level on the display screen 6 is greater than the set noise reduction start threshold. When the indoor noise level is lower than the noise reduction start threshold, the system automatically disables the noise reduction function, which helps save energy consumption.

[0074] S2, Modeling the secondary path of the secondary speaker 7 and the error microphone 5 to obtain secondary path model parameters; including:

[0075] Step 21: Control the secondary speaker 7 to emit sound, and use the LMS algorithm to model the secondary path between the secondary speaker 7 and the error microphone 5; calculate the secondary path error value;

[0076] Step 22: Repeat step S21 to obtain multiple channel error values ​​through multiple iterations. Take the channel error value with the smallest value as the secondary path model parameter and store it in the controller.

[0077] S3, based on the secondary path model parameters, performs noise reduction via an FPGA as the controller, thereby controlling indoor noise, including:

[0078] S31: The outdoor raw noise signal and indoor noise signal, which are collected in real time through multiple reference microphones and multiple error microphones, are respectively used as the input signal and error signal of the controller;

[0079] S32: The noise reduction algorithm is executed via FPGA based on the secondary path model parameters, wherein the noise reduction algorithm includes:

[0080] The adaptive filter coefficients are updated according to formulas (12) and (13) to generate an output signal with the same amplitude but opposite phase as the noise signal; where:

[0081] w j (n+1)=w j (n)-μ(n)e m (n)x′ mj (n) (12);

[0083]

[0084] In equation (12), m represents the error microphone label of the maximum error signal; j represents the secondary sound source label; x′ mj (n) represents the estimated secondary path transfer function of the reference signal after passing through the secondary sound source corresponding to the maximum error microphone. The filtered output after; e m (n) represents the signal vector with the largest error signal amplitude among the K error microphones; w j (n+1) and w j (n) represents the weight coefficient matrix of the transverse FIR adaptive filter; in equation (13), α is used to adjust the problem of large step size caused by the small inner product value of the input signal; β is used to adjust the value of the tanh function to ensure that it is between 0 and 1; x(n) represents the vector form of the output signal of the reference microphone at time n.

[0085] Specifically, in this embodiment, the noise reduction algorithm and its principle are as follows:

[0086] This embodiment is a 1×J×K multi-channel system (1 reference microphone, J secondary sound sources and K error microphones). The reference microphone uses sound insulation technology to eliminate the influence of the secondary sound source output, so the feedback path is not included in the algorithm.

[0087] In active noise control systems, noise control is achieved by minimizing the signal collected by the error microphones using a specific algorithm. The error signal criterion is also known as the cost function. For a multi-channel ANC system with K error microphones, the cost function is defined as J... p .

[0088]

[0089] Where 1≤p≤∞. When p=2, it is the least squares criterion. The multi-channel FxLMS algorithm uses it as the objective function of the control system, that is, to minimize the sum of the squares of the sound pressure amplitudes of the signals collected at each error microphone, so as to minimize the noise field.

[0090] To reduce the computational load of multi-channel systems, this invention employs a change in the cost function J. p The value of p is used to transform the minimum mean square criterion when p=2 into the Minimax criterion when p=∞. Considering the case where p→∞, the ANC problem can be transformed from a minimum noise field energy problem into a minimum noise field amplitude problem, reducing the number of error microphones used in the adaptive filter coefficient update, thereby reducing the system's computational load. In this embodiment, the cost function is defined as:

[0091]

[0092] e m This represents the signal vector with the largest error signal amplitude among the K error microphones. The multi-error ANC algorithm using the Minimax criterion only selects the absolute value of the sample with the largest error signal for each iteration during the adaptive filter coefficient update process. Compared to the multi-error Fxlms algorithm, which uses error signals from all error microphones for coefficient updates, this significantly reduces the computational load. Furthermore, to improve the convergence and steady-state error of the multi-error ANC algorithm using the Minimax criterion, this embodiment combines the input signal normalization with the improved hyperbolic tangent function tanh. This ensures the algorithm's stability even with unknown signals, guaranteeing good steady-state error and fast convergence speed. By setting relevant parameters during the introduction of the tanh function, the problem of poor anti-interference capability in normalization algorithms can be addressed, enabling the algorithm to resist significant noise interference during iterative convergence and ensuring stability. Additionally, the step size changes with each update of the adaptive filter coefficients, essentially making it a variable step-size algorithm, which improves the convergence speed to some extent. The algorithm block diagram is shown below. Figure 3 As shown.

[0093] The adaptive filter uses a transverse FIR filter of length L. The vector form of the reference microphone's output signal at time n is:

[0094] x(n)≡[x(n),x(n-1),…x(n-L+1)] T (3);

[0095] J transverse FIR filters w j The weight coefficient matrix W constitutes the adaptive filter.

[0096] W(n)≡[w1(n), e2(n),…,e J (n)] T (4);

[0097] The output of the j-th secondary sound source at time n can be expressed as:

[0098] y j (n)=x T (n)w j (n) (5);

[0099] in:

[0100] y j (n)≡[y j (n), y j (n-1), ..., y j (n-L+1)] T (6);

[0101] The coefficients of the j-th adaptive filter:

[0102] w j (n)≡[w j (n), w j (n-1), ...w j (n-L+1)] T (7);

[0103] Therefore, for the entire multi-channel system, the output signal at time n is:

[0104] y(n)=x T (n)W(n) (8);

[0105] Assume the secondary path model is also modeled by a transverse FIR filter of length L. Then the secondary path model S is expressed as:

[0106]

[0107] Among them, s kj The secondary path model from the j-th secondary sound source to the k-th error microphone is represented as:

[0108]

[0109] Then the signal e collected by the k-th error microphone k Represented as:

[0110]

[0111] The control filter coefficients are updated according to the Minimax criterion, and a normalization algorithm is introduced to obtain the adaptive filter coefficient update formula (12) for the ANC algorithm:

[0112] w j (n+1)=w j (n)-μ(n)e m (n)x′ mj (n) (12);

[0113]

[0114] In equation (12), m represents the error microphone label of the maximum error signal, and j represents the secondary sound source label; x′ mj (n) represents the secondary path transfer function estimated by the secondary loudspeaker corresponding to the maximum error microphone after the reference signal passes through it. The filtered output is expressed as:

[0115] x′ mj (n)=[x′ mj (n), x′ mj (n-1), ..., x′ mj (n-L+1)] (14);

[0116] in:

[0117]

[0118] In equation (13), α is used to adjust the large step size caused by the small inner product of the input signal; β is used to adjust the value of the tanh function to ensure that it is between 0 and 1.

[0119] This embodiment implements the filter in the algorithm using a serial filter based on dual-port RAM. The entire filter module consists of a read / write control module, dual-port RAM, coefficient ROM, and a multiplier. The dual-port RAM is used for buffering input data, and the coefficient ROM is used for storing filter coefficients. The read / write control module generates read / write address counts to control the filtering process. This method can achieve the highest possible filter order, enabling the noise reduction system to calculate higher-order filters for noise control output and achieve better noise reduction results.

[0120] S33: The output signal is sent by driving multiple secondary speakers (7) and superimposed on the sound wave at the error microphone to achieve noise reduction.

[0121] This embodiment also provides a method for experimental verification using an FPGA platform, the details of which are as follows:

[0122] 1) Hardware Part

[0123] The hardware structure block diagram of the active noise control system is as follows: Figure 5As shown, the I2C communication protocol registers of the ADC and DAC chips are configured using an STM32 microcontroller, setting the ADC and DAC chips as slaves and the data transmission format to I2S. A reference microphone collects the signal from the noise source, and an error microphone array collects the error signal at the noise cancellation point. This signal is converted into a digital signal by the ADC chip and enters the FPGA processing chip through the I2S audio interface. The FPGA processor processes the digital signal to generate an output signal with specific characteristics, which is then converted into an analog signal by the DAC. After power amplification, this signal drives a secondary sound source to emit sound to cancel out the noise at the noise cancellation point.

[0124] 2) Software component

[0125] The ANC algorithm in the software portion of this invention primarily employs a normalization method improved by combining the Minimax criterion with the tanh function. The specific program content of the controller includes:

[0126] (1) I2S audio interface: The ADC and DAC chips are configured as slaves using an STM32 microcontroller, and data transmission is in standard I2S format. The FPGA is used as the master to generate input and output I2S clock signals and perform timing constraints to ensure accurate communication.

[0127] (2) ANC algorithm: The key implementation steps of the algorithm are as follows: update the adaptive filter coefficients.

[0128] w j (n+1)=w j (n)-μ(n)e m (n)x' mj (n) (12);

[0129]

[0130] ROM core: Used to store the coefficients of the fixed filter, i.e. the secondary path estimation transfer function.

[0131] Dual-port RAM core: used for buffering the input data of the fixed filter.

[0132] (5) Data acquisition from the reference microphone and error microphone is used to calculate and display the indoor and outdoor noise levels, respectively.

[0133] The program structure diagram is as follows: Figure 6 As shown.

[0134] Example

[0135] See Figure 1 and Figure 2This invention, based on the designed window model, compares the actual noise reduction effect of the disclosed algorithm with the standard multi-channel Fxlms algorithm, and compares the noise reduction curves at the error microphone. White noise in the 200–2000Hz and 200–1200Hz frequency bands is used as the noise source for testing, and a relevant channel model is established. Before the noise reduction function starts, noise is emitted from the noise source, collected by the reference microphone, and used as the controller input. After processing by the algorithm within the controller, it drives the secondary sound sources on both sides of the opening to emit sounds with certain characteristics, thus reducing noise at the error microphone. Data before and after noise reduction at the error microphone for both algorithms are collected for comparison of the noise reduction effect.

[0136] Figure 7(a) shows the noise reduction effect diagrams and curves at error microphones numbered 1-4 based on two algorithms and a window model; Figure 7(b) shows the noise reduction effect diagrams and curves at error microphones numbered 5-8 based on two algorithms and a window model. It can be seen that the algorithm disclosed in this invention can effectively control outdoor noise propagation, and the noise reduction effect at some noise reduction points is superior to the multi-channel Fxlms algorithm. The algorithm disclosed in this invention significantly reduces the computational load of multi-channel system algorithms, providing the possibility of expanding channels for more complex application scenarios.

[0137] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A window noise reduction system based on a multi-error ANC algorithm under the Minimax criterion, characterized by, The application relates to a multi-channel active noise control system, comprising: a window body (1) and a multi-channel active noise control system, wherein: the multi-channel active noise control system comprises: a plurality of reference microphones (4) located outside a wall body where the window body (1) is located and uniformly distributed on both sides of a window of the window body (1) for collecting original noise signals outside; a plurality of secondary loudspeakers (7) located indoors and uniformly distributed on both sides of the window of the window body (1) for controlled output of anti-noise signals to reduce noise from indoors; a plurality of error microphones (5) located indoors and fixed on an error microphone fixing frame (8); the plurality of error microphones (5) are uniformly distributed in front of the plurality of secondary loudspeakers (7) and used for collecting noise signals indoors; and a controller which carries out noise control indoors by implementing a multi-error ANC algorithm based on a Minimax criterion; the multi-error ANC algorithm based on the Minimax criterion adopts normalization based on a Minimax criterion combined with a tanh function as a control algorithm.

2. The window noise reduction system based on the multi-error ANC algorithm under the Minimax criterion of claim 1, wherein, The controller is an FPGA processor, an STM32 is used to complete configuration of an I2C communication protocol register of an ADC and a DAC chip, the ADC and the DAC chip are set as slaves, a data transmission format is set as an I2S format, the FPGA is set as a host, input and output I2S clock signals are generated, and timing constraints are performed; the reference microphones collect signals of noise sources, the error microphone array collects error signals at noise cancellation points, the signals are converted into digital signals by an ADC chip and input into the FPGA processor through an I2S audio interface, the FPGA processor generates output signals with certain characteristics by processing the digital signals, the output signals are converted into analog signals by a DAC, the analog signals drive the secondary sound sources to emit sound to cancel noise at the noise cancellation points after power amplification; a ROM core is used for storing coefficients of a fixed filter, namely an estimated transfer function of a secondary path; and a dual-port RAM core is used for buffering input data of the fixed filter.

3. The window noise reduction system based on the multi-error ANC algorithm under the Minimax criterion of claim 1, wherein, The application further comprises a remote controller (2) which is used for controlling opening or closing of the multi-channel active noise control system.

4. The window noise reduction system based on the multi-error ANC algorithm under the Minimax criterion of claim 1, wherein, The application further comprises a power indicator light (3) which is used for displaying a working state of the multi-channel active noise control system; when the multi-channel active noise control system is in an open state, the power indicator light (3) is always on.

5. The window noise reduction system based on the multi-error ANC algorithm under the Minimax criterion of claim 1, wherein, The application further comprises a display screen (6) which is used for displaying indoor and outdoor noise levels respectively; a plurality of outdoor original noise signal values are collected by the plurality of reference microphones (4), an average value of the plurality of outdoor original noise signal values is taken as an outdoor noise level and displayed on the display screen (6); a plurality of indoor noise signal values are collected by the plurality of error microphones, an average value of the plurality of indoor noise signal values is taken as an indoor noise level and displayed on the display screen (6).

6. A method for window noise reduction based on a multi-error ANC algorithm under Minimax criterion, implemented by the noise reduction system according to any one of claims 1-5, characterized in that, The application further comprises the following steps: S1, detecting an environmental noise level; S2, modeling a secondary path of the secondary loudspeakers (7) and the error microphones (5) to obtain secondary path model parameters; S3, performing noise reduction by the controller based on the secondary path model parameter, thereby controlling noise in the room.

7. The window noise reduction method of claim 6, wherein, The S1 includes: S11, collecting outdoor original noise signals by multiple reference microphones (4) to obtain a first data set, and sending the first data set to the controller for outdoor noise average value calculation and display as an outdoor noise level, which serves as a noise reduction reference; S12: Collecting indoor noise signals by multiple error microphones (5) to obtain a second data set, and sending the second data set to the controller for indoor noise average value calculation and display as an indoor noise level, which serves as a noise reduction system start condition; S13: Setting a noise reduction start threshold in the active noise control system, and starting noise reduction when the indoor noise level is detected to be greater than the noise reduction start threshold.

8. The window noise reduction method of claim 7, wherein, The S2 includes: Step 21: Control the secondary speaker (7) to make sound, model the secondary path of the secondary speaker (7) and the error microphone (5) using the LMS algorithm; Calculate the secondary path error value; Step 22: Repeat step S21 to obtain multiple channel error values through multiple iterations, and take the minimum channel error value as the secondary path model parameter and store it in the controller.

9. The window noise reduction method of claim 8, wherein, The S3 includes: S31: Collecting outdoor original noise signals and indoor noise signals in real time by multiple reference microphones and multiple error microphones as input signals and error signals of the controller, respectively; S32: Based on the secondary path model parameter, the FPGA executes the multi-error ANC algorithm based on the Minimax criterion; S33: Drive multiple secondary speakers (7) to send output signals, and the sound waves at the error microphone are superimposed on each other, thereby achieving the effect of noise reduction.

10. The window noise reduction method of claim 9, wherein, The multi-error ANC algorithm based on the Minimax criterion in S32 includes: Update the adaptive filter coefficients according to formulas (12) and (13) to generate an output signal with the same amplitude and opposite phase as the noise signal; Wherein: w j (n+1) = w j (n) - μ(n)e m (n)x′ mj (n) (12) (13)； wherein, m in equation (12) represents the error microphone index of the maximum error signal; j represents the secondary sound source index; x' represents the output of the secondary path adaptive filter; x represents the output of the secondary path filter; e represents the error signal; and x represents the output of the secondary path filter. mj (n) represents the estimated secondary path transfer function of the reference signal through the secondary sound source corresponding to the maximum error microphone (n) represents the output of the secondary path filter; e represents the error signal; and x represents the output of the secondary path filter. m (n) represents the signal vector with the maximum error signal amplitude among the K error microphones; w represents the weight coefficient matrix of the transverse FIR filter adaptive filter; and x represents the output of the secondary path filter. j (n) represents the signal vector with the maximum error signal amplitude among the K error microphones; w represents the weight coefficient matrix of the transverse FIR filter adaptive filter; and x represents the output of the secondary path filter. j (n) represents the signal vector with the maximum error signal amplitude among the K error microphones; w represents the weight coefficient matrix of the transverse FIR filter adaptive filter; and x represents the output of the secondary path filter.

Images