Adaptive noise reduction method and system based on FXLMS algorithm, storage medium and device

Abstract

This invention discloses an adaptive noise reduction method based on the FXLMS algorithm. The method includes: S1: acquiring a reference signal containing original noise; S2: processing the reference signal based on a transverse FIR adaptive filter to generate a secondary signal to cancel the original noise; S3: converting the secondary signal into a sound wave, and superimposing it with the original noise in space to generate a noise residual signal; S4: adjusting the weight coefficients of the transverse FIR adaptive filter based on the noise residual signal using the FXLMS algorithm; S5: applying the updated weight coefficients to the transverse FIR adaptive filter, and returning to step S2, iterating continuously until the noise residual signal meets a preset convergence condition. This invention can not only efficiently suppress low-frequency noise that is difficult to handle with traditional passive noise reduction, but also effectively compensate for the influence of the actual acoustic path using the FXLMS algorithm, thereby significantly improving the efficiency of the noise reduction method.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to an adaptive noise reduction method, system, storage medium, and device based on the FXLMS algorithm. Background Technology

[0002] Noise pollution is prevalent in many scenarios, including industrial production, transportation, and daily life, seriously affecting people's physical and mental health and work efficiency. Traditional noise control technologies mainly employ passive noise reduction methods, such as using sound-insulating materials, sound-absorbing structures, or personal protective earmuffs. These passive methods block the propagation path of noise by absorbing or isolating it through physical barriers.

[0003] However, passive noise reduction methods have very limited effectiveness in suppressing low-frequency noise (such as engine roar and air conditioning compressor vibration). This is because low-frequency sound waves have long wavelengths and strong diffraction capabilities, allowing them to easily bypass most physical obstacles, making it difficult for traditional sound insulation and sound absorption materials to effectively block them.

[0004] To overcome the limitations of passive noise reduction, active noise control technology has emerged. Its core idea is to use sound to cancel sound, that is, to generate a secondary sound wave with the same amplitude but opposite phase as the original noise through an electroacoustic system, so that when the two meet in space, they will interfere destructively, thereby significantly reducing the energy of the residual noise.

[0005] The key to achieving active noise control lies in how to accurately and rapidly generate the required secondary sound waves. This requires a control algorithm capable of tracking changes in noise characteristics in real time and automatically adjusting accordingly. While existing adaptive filtering algorithms can achieve adaptive adjustment, their convergence speed and stability are limited when dealing with active noise cancellation systems containing secondary sound channels, potentially leading to poor noise reduction results or even system divergence. Therefore, designing an adaptive noise reduction method that can effectively compensate for the influence of secondary sound channels, achieve fast convergence, and operate stably is a pressing technical problem to be solved in this field. Summary of the Invention

[0006] Therefore, it is necessary to propose an adaptive noise reduction method based on the FXLMS algorithm to address the above problems.

[0007] An adaptive noise reduction method based on the FXLMS algorithm, the method comprising the following steps: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2, continuing the iteration until the noise residual signal meets the preset convergence condition.

[0008] In the above scheme, the step of processing the reference signal based on a transverse FIR adaptive filter to generate a secondary signal for canceling the original noise specifically includes: in, For secondary signals used to cancel the original noise, For reference signal, This is the weight coefficient vector.

[0009] In the above scheme, the step of converting the secondary signal into a sound wave and then superimposing it with the original noise in space to generate a noise residual signal specifically includes: The noise residual signal is determined using the following formula: in, For noise residual signal, For secondary signals used to cancel the original noise, The original noise signal before noise reduction was applied. For secondary channel impulse response, This represents convolution.

[0010] In the above scheme, the step of converting the secondary signal into a sound wave and then superimposing it with the original noise in space to generate a noise residual signal further includes: The actual sound field signal is collected using a microphone located in the area to be noise-reduced. The actual sound field signal is used as the noise residual signal. This is used for updating the weight coefficients in step S4.

[0011] In the above scheme, adjusting the weight coefficients of the transverse FIR adaptive filter using the FXLMS algorithm based on the noise residual signal specifically includes: Reference signal The signal is obtained by filtering the estimation model through a secondary channel. ; Update the weights of the transverse FIR adaptive filter according to the following formula: in, For the updated weight coefficient vector, For reference signal, To estimate the filtered signal of the model, For weight coefficient vector, For noise residual signal, This is the convergence step size factor.

[0012] In the above scheme, the convergence step size factor The order N of the transverse FIR adaptive filter is configured according to the number of frequency bands contained in the noise to be canceled.

[0013] In the above scheme, the preset convergence condition includes at least one of the following: The energy or amplitude of the residual signal drops below a preset threshold; the change in the weight coefficient vector of the transverse FIR adaptive filter is lower than a preset tolerance; and the iteration process reaches a preset duration or number of iterations.

[0014] This application also proposes an adaptive noise reduction system based on the FXLMS algorithm, the system comprising: a signal acquisition unit, an adaptive filtering unit, a noise residual signal determination unit, and an update unit; The signal acquisition unit is used to acquire a reference signal containing the original noise; The adaptive filtering unit adopts a transverse FIR structure to process the reference signal and generate a secondary signal to cancel the original noise. The noise residual signal determination unit is used to convert the secondary signal into a sound wave, superimpose it with the original noise in space, and collect the noise residual signal generated after superposition. The updating unit is used to adjust the weight coefficients of the adaptive filtering unit based on the noise residual signal using the FXLMS algorithm, and apply the updated weight coefficients to the adaptive filtering unit; wherein the adaptive noise reduction system based on the FXLMS algorithm is configured to perform iterative operations until the noise residual signal meets the preset convergence condition.

[0015] This application also proposes a readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2, continuing the iteration until the noise residual signal meets the preset convergence condition.

[0016] This application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor in the following steps: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2, continuing the iteration until the noise residual signal meets the preset convergence condition.

[0017] The embodiments of this invention offer the following advantages: By employing adaptive closed-loop control based on the FXLMS algorithm, this invention can effectively generate secondary sound waves that are precisely out of phase with the original low-frequency noise. Utilizing the principle of acoustic wave interference cancellation, it directly cancels the low-frequency noise energy within the target area, compensating for the shortcomings of traditional passive noise reduction methods in the low-frequency band. Simultaneously, by utilizing a transverse FIR adaptive filter, it can automatically adjust the filter weight coefficients online and in real-time based on the noise residual signal, adapting to unknown or time-varying noise characteristics and improving noise reduction performance in complex acoustic environments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] in: Figure 1 This is a schematic diagram of an adaptive noise reduction method based on the FXLMS algorithm in one embodiment; Figure 2 This is a schematic diagram of the structure of a transverse FIR adaptive filter in one embodiment; Figure 3 This is a time-domain plot of the noise residual of a filter with order 10 versus the noise signal and control signal under a single-frequency noise environment in one embodiment. Figure 4 This is a time-domain plot of the noise residual versus the noise signal and control signal for a filter of order 50 under a single-frequency noise environment in one embodiment. Figure 5 This is a time-domain plot of the noise residual with respect to the noise signal and control signal under a single-frequency noise environment with a step size factor of 0.001, as shown in one embodiment. Figure 6 This is a time-domain plot of the noise residual with respect to the noise signal and control signal under a single-frequency noise environment with a step size factor of 0.003, as shown in one embodiment. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention; however, it will be apparent to those skilled in the art that the invention may be practiced without one or more of these details; in other instances, certain technical features well-known in the art have not been described in order to avoid confusion with the invention. It should be understood that the invention can be practiced in different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided to make the disclosure thorough and complete and to fully convey the scope of the invention to those skilled in the art.

[0022] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms, unless the context clearly indicates otherwise. The terms “comprising” and / or “including,” when used in this specification, identify the presence of said features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0023] With the rapid development of industrial civilization and urbanization, noise pollution has become a significant environmental problem affecting people's daily lives, work, and health. Low-frequency noise, in particular, suffers from long wavelengths and strong diffraction capabilities, making traditional passive noise reduction techniques such as sound insulation and sound-absorbing materials ineffective and prone to drawbacks such as bulkiness and difficulty in optimizing for specific frequencies. Therefore, active noise control (ANC) technology has emerged. Its basic principle is to generate a secondary sound wave with the same amplitude but opposite phase as the original noise, causing them to superimpose and interfere in space, thereby canceling out sound energy. Among numerous ANC algorithms, the filter-X Least Mean Square (FXLMS) algorithm based on the least mean square algorithm has become the mainstream in research and application due to its simple principle and ease of implementation. However, the traditional FXLMS algorithm still faces several serious challenges in practical engineering applications: the ideal FXLMS algorithm model assumes that the acoustic path from the secondary speaker to the error microphone is linear and time-invariant. In reality, however, this channel characteristic can change due to temperature variations, equipment aging, or small changes in the relative positions of the speaker and microphone, leading to decreased convergence performance or even divergence, and insufficient system robustness. Furthermore, the fixed step size factor μ in the algorithm cannot balance convergence speed and steady-state error. A larger step size factor can accelerate the initial convergence speed, but it leads to increased steady-state error and may even cause system instability when secondary channels change; a smaller step size factor can achieve a smaller steady-state error, but the convergence speed is slow and it is difficult to quickly track non-stationary noise. Finally, when faced with real-world industrial or traffic noise with complex frequency components and uneven energy distribution, filters with fixed parameters struggle to achieve optimal noise reduction performance across all frequency bands. The algorithm lacks the ability to perceive noise characteristics and cannot adaptively adjust, resulting in unsatisfactory noise reduction performance for sudden or broadband noise.

[0024] Therefore, there is an urgent need in this field for an improved adaptive noise reduction method based on the FXLMS algorithm, which can effectively improve the robustness to the time-varying characteristics of the secondary channel, intelligently balance the convergence speed and steady-state accuracy, and enhance the adaptability to complex noise environments, so as to meet the high-standard practical noise reduction application requirements.

[0025] To fully understand the present invention, a detailed structure will be presented in the following description in order to illustrate the technical solution proposed by the present invention; optional embodiments of the present invention are described in detail below, however, in addition to these detailed descriptions, the present invention may have other embodiments.

[0026] like Figure 1 As shown, in one embodiment, an adaptive noise reduction method based on the FXLMS algorithm is provided. This adaptive noise reduction method based on the FXLMS algorithm includes steps S1 to S5, which are detailed below: S1: Acquire a reference signal containing the original noise; This step provides the necessary data input for the entire adaptive noise reduction system and is the prerequisite and foundation for achieving noise cancellation. Only by acquiring real-time samples of the original noise can a targeted countermeasure signal be generated.

[0027] Preferably, the microphone is deployed in an area near the original noise source and unaffected by secondary signal interference, with a sampling rate of 8000Hz and a sampling precision of 16 bits. This achieves accurate acquisition of low-frequency noise in the 100Hz-500Hz range, with a signal-to-noise ratio ≥30dB, providing a high-fidelity raw data foundation for the generation of subsequent secondary signals, and an acquisition delay ≤1ms, meeting the timing requirements of real-time noise reduction.

[0028] S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; By utilizing the inherent linear phase and absolute stability of FIR filters, it is ensured that the generated secondary signal will not introduce phase distortion or cause system oscillation, thus avoiding the risk of oscillation due to feedback. This step, based on the current control algorithm model, calculates in real time an electrical signal with a specific relationship in amplitude and phase with the reference signal, serving as the command to drive the loudspeaker; it is a crucial computational step in achieving sound cancellation.

[0029] Preferably, for noise with different frequencies, the transverse FIR adaptive filter can achieve accurate secondary signal generation. For single-frequency 500Hz noise, the phase difference between the generated secondary signal and the original noise is controlled within the range of 175°-185°, with an amplitude error ≤2%. For multi-frequency mixed noise, it can simultaneously generate anti-phase sound waves corresponding to multiple frequencies, without frequency crosstalk, and the filter has no risk of feedback oscillation, achieving an operational stability of 99.9%.

[0030] like Figure 2 The diagram shows the structure of a transverse FIR adaptive filter, which consists of a series of weighted tapped delay lines. The input signal... Passing sequentially through the unit delay unit z 1 The weighting coefficients corresponding to each tap After multiplying and summing, the output signal is obtained. Its weighting coefficients can be adjusted online according to adaptive algorithms (such as LMS, FXLMS, etc.), so that the filter can dynamically track the changes in the statistical characteristics of the input signal and achieve optimal filtering or noise cancellation for a specific signal.

[0031] In some embodiments, the reference signal is processed based on a transverse FIR adaptive filter to generate a secondary signal for canceling the original noise, specifically including: in, For secondary signals used to cancel out the original noise, For reference signal, This is the weight coefficient vector.

[0032] This embodiment explicitly defines the core computational process for generating the secondary signal using a transverse FIR adaptive filter through mathematical formulas. It leverages the inherent absolute stability of FIR filters to fundamentally avoid the risk of system oscillations, providing a reliable foundation for the entire adaptive noise reduction loop. Simultaneously, this structure transforms the complex acoustic cancellation problem into a problem involving the weighting coefficient vector. The dynamic optimization problem enables the system to adjust these weight coefficients in real time and automatically through the subsequent FXLMS algorithm, thereby accurately generating an antiphase sound wave that matches the amplitude and phase characteristics of the variable noise. This not only achieves a leap from static filtering to dynamic adaptive tracking, but also, due to the regularity and parallelism of its computational process, makes it possible to implement the algorithm efficiently and in real time on low-cost digital signal processors, laying the foundation for the engineering of the entire noise reduction scheme.

[0033] S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; This step represents the transition from digital control to physical action. It transforms the electrical signal calculated in the previous step into a real sound wave, i.e., a secondary sound source, via a loudspeaker, and then uses the principle of sound wave interference to superimpose energy with the original noise in space. This is the physical execution process by which the entire method produces the actual noise reduction effect. The generated noise residual signal is the result of the interference, directly reflecting the effectiveness of the noise reduction at the current moment.

[0034] In some embodiments, the secondary signal is converted into a sound wave, which is then spatially superimposed with the original noise to generate a noise residual signal, specifically including: The noise residual signal is determined using the following formula: in, For noise residual signal, For secondary signals used to cancel out the original noise, The original noise signal before noise reduction was applied. For secondary channel impulse response, This represents convolution.

[0035] This embodiment introduces convolution operations and secondary channel impulse responses. It accurately models the entire path from digital signals to physical sound fields, and its technical effectiveness is crucial. It first anchors the algorithm from an ideal digital domain model to the complex physical reality, through... The algorithm accurately predicts the delay, attenuation, and distortion experienced by the secondary sound wave as it propagates from the speaker to the error microphone, enabling it to recognize and compensate for these physical effects in advance. Secondly, the formula... The definition of noise residual signal Explicitly defined as the original noise that is to be eliminated The difference between the actual secondary sound wave and the actual sound wave shaped by the physical path provides the only real and measurable feedback for the entire adaptive system. It is based on this precise error signal that the subsequent FXLMS algorithm can calculate the correct weight update direction and drive convergence. This modeling fundamentally solves the core problem of system performance degradation or even instability caused by ignoring the actual acoustic path, greatly improving the effectiveness and robustness of active noise cancellation algorithms in real-world applications. It is an indispensable key element in achieving high-performance adaptive noise cancellation.

[0036] In some embodiments, converting the secondary signal into a sound wave and spatially superimposing it with the original noise to generate a noise residual signal further includes: The actual sound field signal is collected using a microphone located in the area to be noise-reduced. The actual sound field signal is used as the noise residual signal. This is used for updating the weight coefficients in step S4.

[0037] This embodiment directly acquires the real sound field signal of the target area using an error microphone as the noise residual, constructing a direct feedback path from the physical world to the digital algorithm. Its core effect lies in transforming an open-loop theoretical model into an adaptive closed-loop system with realistic perception capabilities. This design allows the algorithm to adjust filter weights based on the most realistic noise reduction effect, rather than ideal simulation results. This automatically compensates for non-ideal factors such as secondary channel modeling errors and changes in the sound field environment, greatly enhancing the system's robustness, adaptability, and reliability of the final noise reduction effect in complex real-world environments. It ensures that advanced signal processing algorithms can accurately act on the physical sound field, achieving stable and optimal active noise control at the target location.

[0038] Preferably, the digital secondary signal The sound is converted to analog voltage by a DAC, then amplified by a power amplifier to drive a speaker. A secondary sound source emits an inverted sound wave. This sound wave is superimposed on the original noise propagating through the air in the target noise reduction area, such as the earphone canal. An error microphone located in this area collects the superimposed sound pressure in real time, converts it into an electrical signal, and samples it using an ADC to obtain the noise residual signal. This step represents the transition from the digital domain to the physical domain. (Noise residual signal) It directly and quantitatively reflects the noise reduction effect at the current moment, providing crucial feedback information for adaptive optimization.

[0039] S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; The FXLMS algorithm automatically and iteratively corrects the control model, i.e., the filter weight coefficients, by analyzing the relationship between the current noise reduction effect (i.e., the residual signal) and the reference signal. Its key advantage lies in the fact that the algorithm includes compensation for the physical acoustic path (i.e., the secondary channel) from the loudspeaker to the error microphone, thereby ensuring the correctness of the update direction in complex acoustic environments, converging to the optimal or suboptimal cancellation state, and tracking time-varying noise.

[0040] In some embodiments, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm based on the noise residual signal, specifically including: Reference signal The signal is obtained by filtering the estimation model through a secondary channel. ; Update the weights of the transverse FIR adaptive filter according to the following formula: in, For the updated weight coefficient vector, For reference signal, To estimate the filtered signal of the model, For weight coefficient vector, For noise residual signal, This is the convergence step size factor.

[0041] This embodiment pre-filters the reference signal by introducing a secondary channel estimation model to generate a filtered -X signal. It effectively compensates for the phase delay and amplitude distortion of the actual acoustic path, ensuring that the weight vector is always updated along the steepest descent direction of the error hyperplane, thus significantly improving the convergence speed and stability of the algorithm. Simultaneously, by explicitly modeling the secondary channel, it achieves inherent robustness to changes in the speaker-microphone path, enabling adaptive compensation for the effects of temperature drift, device aging, or minor changes in physical position. This mechanism fundamentally solves the core problem of traditional LMS algorithms being prone to divergence in active noise control due to neglecting the characteristics of secondary channels, making it possible for the FXLMS algorithm to achieve reliable and efficient adaptive noise reduction in real physical systems.

[0042] In some embodiments, the convergence step size factor The order N of the transverse FIR adaptive filter is configured according to the number of frequency bands contained in the noise to be canceled.

[0043] In practice, when the step size factor of the FXLMS algorithm remains constant, the filtering performance of the transverse FIR adaptive filter is affected by the filter order. As the order of the transverse FIR adaptive filter increases, the convergence speed of the system also increases.

[0044] When the filter order of the FXLMS algorithm remains constant, the filtering performance of the transverse FIR adaptive filter is affected by the step size factor. As the step size factor increases, the filter weight coefficients in each iteration approach the optimal filter weight coefficients, meaning the convergence speed increases. However, if the step size factor is too large, it will lead to an increase in the filter weight coefficients after each iteration, causing the system to diverge. Therefore, a suitable step size factor should be chosen while ensuring a relatively ideal system convergence speed.

[0045] Preferably, when the noise is single-frequency noise, the convergence step size factor is... The value range is from 0.001 to 0.003; when the noise is dual-frequency noise, the convergence step size factor is... The value range is from 0.0001 to 0.0005; when the noise is three-frequency or four-frequency noise, the convergence step size factor is... The value of L ranges from 0.00005 to 0.0002. When the noise is single-frequency noise, the order N of the transverse FIR adaptive filter ranges from 10 to 50; when the noise is dual-frequency noise, the order N of the transverse FIR adaptive filter ranges from 25 to 120; when the noise is tri-frequency noise, the order N of the transverse FIR adaptive filter ranges from 40 to 200; and when the noise is quad-frequency noise, the value of L ranges from 30 to 160.

[0046] In practice, for the simplest scenario like single-frequency noise, a large step size factor can be used to achieve extremely fast convergence. Simultaneously, because the noise spectrum is singular, a small filter order is sufficient for accurate modeling and generation of the counter-signal. This configuration has low computational cost and high real-time performance, making it ideal for resource-sensitive applications with simple noise environments, such as noise reduction for single-frequency devices. For example: Configuration: Convergence Step Size Factor The value ranges from 0.001 to 0.003; the order N of the transverse FIR adaptive filter ranges from 10 to 50.

[0047] See Figure 3 As shown, in a single-frequency noise environment (a mixture of a 500Hz sine wave and Gaussian white noise), with a fixed step size factor μ = 0.002, when the order of the transverse FIR adaptive filter is set to a relatively low 10th order, the system reaches convergence after approximately 500 iterations, and the energy of the noise residual signal is significantly reduced. See also... Figure 4 As shown, under the same single-frequency noise environment and step size factor setting, the order of the transverse FIR adaptive filter is increased to 50. (Comparison) Figure 3 As can be seen, the system convergence speed is significantly faster, requiring only about 100 iterations to reach convergence, and the noise residual in steady state is lower.

[0048] By comparison Figure 3 and Figure 4 Simulation results show that the filter order in the method of this invention has a direct impact on the system convergence speed and noise reduction accuracy. When the filter order is low (e.g., 10th order), the filter's ability to model the acoustic path is limited, resulting in a slow convergence process (requiring 500 iterations); however, when the order is increased to 50th order, the filter gains sufficient degrees of freedom to accurately approximate the system response, and the convergence speed is improved by about 5 times.

[0049] See Figure 5 As shown, in a single-frequency noise environment, the order of the transverse FIR adaptive filter is fixed at 20. When the step size factor is set to a small 0.001, the system convergence speed is moderate, requiring approximately 300 iterations to reach steady state, and the noise residual signal is smooth and without fluctuations. See [link / reference] Figure 6 As shown, under the same conditions, the step size factor is increased to 0.003. (Comparison) Figure 5 As can be seen, the system convergence speed is significantly accelerated, and it only takes about 200 iterations to reach a steady state, which greatly improves the real-time performance of the noise reduction response.

[0050] By comparison Figure 5 and Figure 6Simulation results show that appropriately increasing the step size factor allows the filter weight coefficients to approach the optimal value in each iteration, thereby shortening the convergence time. Therefore, in burst noise scenarios with extremely high response speed requirements, a larger step size factor can be used to achieve rapid noise reduction, thus filling the gap in traditional passive noise reduction's inability to dynamically adjust for specific frequencies.

[0051] Furthermore, when the noise contains two frequency components, a stronger modeling capability is required. Therefore, the filter order N needs to be increased to capture more complex frequency characteristics. Simultaneously, to maintain stability at both frequencies, the step size factor m must be reduced to avoid update overshoot. This embodiment achieves a good balance between convergence speed and system stability, and is suitable for handling noise sources containing a dominant frequency and significant harmonics or harmonics, such as certain motor or mechanical vibrations. For example, configuring the convergence step size factor... The value ranges from 0.0001 to 0.0005; the order N of the transverse FIR adaptive filter ranges from 25 to 120.

[0052] Furthermore, in broadband or complex spectral noise environments, this embodiment employs a very large filter order N, granting the filter sufficiently high degrees of freedom to accurately approximate the complex inverse acoustic channel response. Simultaneously, a very small step size factor m ensures algorithm stability under complex update directions, effectively preventing system divergence. This demonstrates the robustness and high accuracy of this invention in handling harsh noise environments, making it suitable for complex scenarios such as engine compartments and factory workshops. For example, for three-frequency or four-frequency noise, a convergence step size factor can be configured. The value ranges from 0.00005 to 0.0002; the order N of the transverse FIR adaptive filter ranges from 40 to 200.

[0053] Preferably, the convergence step size factor Replace with a noise residual signal Dynamically changing step size factor The specific formula is as follows: in, The replacement step size factor, For noise residual signal, This is a positive constant used to control the range of step size variation.

[0054] This preferred embodiment resolves the inherent contradiction between convergence speed and steady-state error in a fixed step size factor. In the initial stage of the algorithm, the error... Larger, step size The error is also relatively large, thus accelerating convergence; when approaching steady state, the error... Decrease, step size This also reduces the steady-state imbalance and improves system stability. This embodiment achieves superior overall performance in various noise environments.

[0055] S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2 to continue iterating until the noise residual signal meets the preset convergence condition.

[0056] By setting convergence conditions, this method can enter steady-state operation after achieving satisfactory noise reduction results, or stop iterating when parameters are inappropriate, thus ensuring the controllability, stability, and practicality of noise reduction. The entire iterative process requires no manual intervention, achieving fully automatic and continuous noise reduction optimization.

[0057] In some embodiments, the preset convergence criteria include at least one of the following: The energy or amplitude of the residual signal drops below a preset threshold; the change in the weight coefficient vector of the transverse FIR adaptive filter is lower than the preset tolerance; the iterative process reaches the preset duration or number of iterations.

[0058] Preferably, the adaptive noise reduction method based on the FXLMS algorithm of this invention is integrated with passive noise reduction materials (such as sound-absorbing cotton and sound insulation boards). The adaptive noise reduction method based on the FXLMS algorithm focuses on processing low-frequency noise below 500Hz, while the passive materials are responsible for absorbing mid-to-high-frequency noise. This hybrid approach combines the advantages of both technologies. Active noise reduction effectively solves the low-frequency problem that passive noise reduction struggles to handle, while passive materials compensate for the shortcomings of active noise reduction, such as poor performance and high cost in the high-frequency range. Ultimately, a balanced and efficient noise reduction effect can be achieved across the entire frequency band, while optimizing cost and power consumption.

[0059] This application also proposes an adaptive noise reduction system based on the FXLMS algorithm, the system comprising: a signal acquisition unit, an adaptive filtering unit, a noise residual signal determination unit, and an update unit; The signal acquisition unit is used to acquire a reference signal containing the original noise; The adaptive filtering unit, employing a transverse FIR structure, is used to process the reference signal and generate a secondary signal to cancel out the original noise. The noise residual signal determination unit is used to convert the secondary signal into a sound wave, superimpose it with the original noise in space, and collect the noise residual signal generated after superposition. The update unit is used to adjust the weight coefficients of the adaptive filtering unit based on the noise residual signal using the FXLMS algorithm, and apply the updated weight coefficients to the adaptive filtering unit; wherein, the adaptive noise reduction system based on the FXLMS algorithm is configured to perform iterative operations until the noise residual signal meets the preset convergence condition.

[0060] Preferably, the signal acquisition unit consists of a high-sensitivity reference microphone, a signal conditioning module, and an A / D conversion module. It is deployed within a 3-5m range around the noise source and can acquire signals from -40dBV to +10dBV. It also has anti-electromagnetic interference capability, and the distortion of the acquired signal in an industrial electromagnetic environment is ≤1%.

[0061] Preferably, the adaptive filtering unit uses an FPGA chip as the core computing carrier and has a built-in transverse FIR adaptive filter logic module, which can realize parallel operation of weight coefficients with a computation latency of ≤0.1ms and supports dynamic configuration of filter order from 10 to 200 to meet the computational needs of different noise scenarios.

[0062] Preferably, the noise residual signal determination unit includes a D / A converter, a power amplifier, a secondary speaker, and an error microphone. The frequency response range of the secondary speaker covers 100Hz-500Hz, and the sensitivity is ≥90dB. The error microphone is deployed at the center of the area to be noise-reduced and can provide real-time feedback on the sound field superposition effect with an acquisition delay ≤0.8ms.

[0063] Preferably, the update unit is based on an ARM processor to build an FXLMS algorithm operation module, which can receive residual signals in real time and complete the weight coefficient update. The update frequency can reach 1kHz, and it supports remote configuration of algorithm parameters, which is convenient for adaptation and adjustment in different scenarios.

[0064] Preferably, for personal portable noise cancellation needs, the original solution is optimized for low power consumption. The adaptive filtering unit is moved to the Bluetooth chip, the signal acquisition unit uses a miniature MEMS microphone, and the secondary sound unit is an in-ear miniature speaker. The overall device volume is ≤5cm³, and the power consumption is reduced to below 500mW. At the same time, the algorithm operation logic is optimized, and fixed-point arithmetic is used instead of floating-point arithmetic to reduce the processor's computational load.

[0065] The system meets the usage requirements in scenarios such as low-frequency noise in vehicles, roaring of industrial equipment, and vibration of air conditioning compressors.

[0066] This application also proposes a readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2 to continue iterating until the noise residual signal meets the preset convergence condition.

[0067] This application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor in the following steps: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2 to continue iterating until the noise residual signal meets the preset convergence condition.

[0068] Those skilled in the art will understand that implementing all or part of the processes in the above embodiments can be accomplished by instructing related hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0069] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0070] The embodiments described above are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application's patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. The embodiments disclosed above are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made according to the claims of this invention are still within the scope of this invention.

Claims

1. An adaptive noise cancellation method based on FXLMS algorithm, characterized in that, The method includes: S1: Acquire a reference signal containing the original noise; S2: Based on the transverse FIR adaptive filter, the reference signal is processed to generate a secondary signal used to cancel the original noise; S3: Convert the secondary signal into a sound wave, and superimpose it with the original noise in space to generate a noise residual signal; S4: Based on the noise residual signal, the weight coefficients of the transverse FIR adaptive filter are adjusted using the FXLMS algorithm; S5: Apply the updated weight coefficients to the transverse FIR adaptive filter and return to step S2, continuing the iteration until the noise residual signal meets the preset convergence condition.

2. The method of claim 1, wherein the FXLMS algorithm is based on, The process of processing the reference signal using a transverse FIR adaptive filter to generate a secondary signal for canceling the original noise specifically includes: wherein is a secondary signal for canceling the original noise, is a reference signal, is a weight coefficient vector.

3. The method of claim 2, wherein the FXLMS algorithm is based on, The step of converting the secondary signal into a sound wave and then superimposing it with the original noise in space to generate a noise residual signal specifically includes: The noise residual signal is determined using the following formula: wherein is a noise residual signal, is a secondary signal for cancelling the original noise, is an original noise signal before applying noise reduction, is a secondary channel impulse response, denotes a convolution.

4. The method of claim 2, wherein the FXLMS algorithm is based on, The step of converting the secondary signal into a sound wave and then superimposing it with the original noise in space to generate a noise residual signal further includes: The actual sound field signal is collected using a microphone located in the area to be noise-reduced. using the actual sound field signal as a noise residual signal for the weight coefficient update of step S4.

5. The method of adaptive noise cancellation based on the FXLMS algorithm according to any one of claims 3 or 4, characterized in that, The step of adjusting the weight coefficients of the transverse FIR adaptive filter based on the noise residual signal using the FXLMS algorithm specifically includes: reference signals filtering through an estimation model of a secondary channel to obtain an estimation model filtered signal ; Update the weights of the transverse FIR adaptive filter according to the following formula: in, For the updated weight coefficient vector, For reference signal, To estimate the filtered signal of the model, For weight coefficient vector, For noise residual signal, This is the convergence step size factor.

6. The adaptive noise reduction method based on the FXLMS algorithm according to claim 5, characterized in that, The convergence step size factor The order N of the transverse FIR adaptive filter is configured according to the number of frequency bands contained in the noise to be canceled.

7. The adaptive noise reduction method based on the FXLMS algorithm according to claim 1, characterized in that, The preset convergence condition includes at least one of the following: The energy or amplitude of the residual signal drops below a preset threshold; the change in the weight coefficient vector of the transverse FIR adaptive filter is lower than a preset tolerance; and the iteration process reaches a preset duration or number of iterations.

8. An adaptive noise reduction system based on the FXLMS algorithm, characterized in that, The system includes: a signal acquisition unit, an adaptive filtering unit, a noise residual signal determination unit, and an update unit; The signal acquisition unit is used to acquire a reference signal containing the original noise; The adaptive filtering unit adopts a transverse FIR structure to process the reference signal and generate a secondary signal to cancel the original noise. The noise residual signal determination unit is used to convert the secondary signal into a sound wave, superimpose it with the original noise in space, and collect the noise residual signal generated after superposition. The updating unit is used to adjust the weight coefficients of the adaptive filtering unit based on the noise residual signal using the FXLMS algorithm, and apply the updated weight coefficients to the adaptive filtering unit; wherein the adaptive noise reduction system based on the FXLMS algorithm is configured to perform iterative operations until the noise residual signal meets the preset convergence condition.

9. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, the processor performs the steps of the method as described in any one of claims 1 to 7.

10. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images