A cascaded predictive active noise cancellation system

Abstract

This invention discloses a cascaded predictive active noise reduction system for pipeline noise reduction. It includes multiple microphones spaced apart along the length of the pipeline; a loudspeaker is positioned between two adjacent microphones, each loudspeaker connected to the pipeline via a separate secondary channel; an analog-to-digital converter (ADC) receives analog sound signals collected by the microphones, converts them into digital sound signals, and sends them to a digital processing unit (DMU); the DMU receives the digital sound signals from the ADC, executes a noise reduction algorithm based on the digital sound signals corresponding to two adjacent microphones, obtains the noise-reduced digital signal for the loudspeaker between the two adjacent microphones, and sends it to an analog-to-digital converter (A / D converter); the A / D converter receives the noise-reduced digital signal from the DMU, ​​converts it into a noise-reduced analog signal, and sends it to the loudspeaker, causing the loudspeaker to emit a corresponding noise-reduced wave to achieve active noise reduction. This invention achieves effective noise reduction in pipelines.

Description

Technical Field

[0001] This invention belongs to the field of active noise reduction technology, specifically relating to a cascaded predictive active noise reduction system. Background Technology

[0002] From a physiological point of view, any sound that interferes with people's rest, study, and work, or interferes with the sounds people want to hear—that is, unwanted sounds—is collectively referred to as noise. When noise has adverse effects on people and the surrounding environment, it constitutes noise pollution.

[0003] Currently, there are three methods for noise reduction: reducing noise at the source, intercepting noise during its propagation, and reducing noise at the ear. Reducing noise at the source involves industries and transportation sectors using low-noise production equipment and improving production processes, or altering the movement of the noise source (e.g., using damping and vibration isolation to reduce the vibration of solid sound-emitting bodies). Intercepting noise during its propagation involves controlling the noise's propagation and changing the path of the noise already emitted by the source, such as using sound absorption, sound insulation, sound barriers, and vibration isolation. For noise protection of the recipient or their organs, when measures cannot be taken at the source or propagation path, or when existing acoustic measures are insufficient, protective measures for the recipient or their organs are necessary. For example, workers with long-term occupational noise exposure can wear earplugs, earmuffs, or helmets.

[0004] Noise reduction can be divided into two types: active noise reduction and passive noise reduction. The methods mentioned above for controlling sound propagation and protecting the receiving organs are mostly passive noise reduction methods, primarily involving the absorption or reflection of noise to reduce its reach to the ear. Active noise reduction, on the other hand, reduces noise by emitting sound waves with the same frequency and amplitude but opposite phase to the noise, thus achieving phase cancellation through interference. For example... Figure 1 As shown, in an active noise cancellation system, a microphone collects noise, the processor inverts the noise, and then emits it through a speaker at an appropriate time, mixing it with the noise. Theoretically, the superimposed sound can completely eliminate the noise.

[0005] Currently, patents for open-space active noise cancellation applications in propagation paths are mostly concentrated in automotive NVH and range hood systems. For example, patent CN105513585A provides an active noise cancellation method and system for automobiles, which uses speakers placed in the ABCD pillars of the car, pressure sensors to sense the center position, and active noise cancellation at the center position. Another example is patent CN108916949A, which provides a low-damping semi-enclosed three-dimensional spatial sound field noise cancellation range hood and noise cancellation method, which divides the sound wave into multiple sub-band signals and uses corresponding speakers for sound field noise cancellation. However, the above-mentioned technologies are all used in specific scenarios, are strongly dependent on the structure, and have poor universality. Therefore, this patent will propose a solution for active noise cancellation in general ducts, employing multi-level reference / error microphones, the FxLMS algorithm, and a time series prediction algorithm. Summary of the Invention

[0006] The purpose of this invention is to provide a cascaded predictive active noise reduction system to achieve effective noise reduction of pipelines.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] A cascaded predictive active noise cancellation system is applied to pipeline noise reduction. The cascaded predictive active noise cancellation system includes a microphone, a speaker, an analog-to-digital converter, a digital-to-analog converter, and a digital processing unit, wherein:

[0009] The microphone is multiple, and the multiple microphones are arranged at intervals along the length of the pipe inside the pipe;

[0010] The speaker has multiple speakers, with one speaker positioned between two adjacent microphones, and each speaker is connected to the pipe through a separate secondary channel;

[0011] The analog-to-digital converter is used to receive the analog sound signal collected by the microphone, convert the analog sound signal into a digital sound signal, and then send it to the digital processing unit;

[0012] The digital processing unit is used to receive the audio digital signal sent by the analog-to-digital converter, execute a noise reduction algorithm based on the audio digital signals corresponding to two adjacent microphones, obtain the noise reduction digital signal of the speaker between the two adjacent microphones, and send it to the digital-to-analog converter.

[0013] The digital-to-analog converter is used to receive the noise-reduced digital signal sent by the digital processing unit, convert the noise-reduced digital signal into a noise-reduced analog signal, and send it to the speaker so that the speaker emits a corresponding noise-reduced wave to achieve active noise reduction.

[0014] Several alternative methods are provided below, but they are not intended as additional limitations on the overall solution above. They are merely further additions or optimizations. Provided there are no technical or logical contradictions, each alternative method can be combined individually with respect to the overall solution above, or multiple alternative methods can be combined with each other.

[0015] Preferably, the microphone is positioned at the center of the pipe cross-section and faces the noise source.

[0016] Preferably, the angle between the secondary channel and the duct is less than 90° in the direction of the noise source.

[0017] Preferably, the mixing point of the noise wave emitted by the loudspeaker and the noise wave in the pipe on the central axis of the pipe is called the mixing point. Two adjacent microphones and the loudspeaker between the two adjacent microphones are used as a noise reduction unit, and the microphone closer to the noise source in the noise reduction unit is used as the reference microphone, and the other microphone in the noise reduction unit is used as the error microphone.

[0018] The positions of the speaker and the reference microphone in the same noise reduction unit satisfy the following relationship:

[0019]

[0020] In the formula, L micm L is the distance between the reference microphone and the mixing point. spkm The distance between the speaker and the mixing point is Vs, where Vs is the speed of sound and t is the distance between the speaker and the mixing point. da The digital-to-analog conversion time calibrated for the digital-to-analog converter, t ad The calibrated analog-to-digital conversion time t for the analog-to-digital converter p The time required for the digital processing unit to execute the noise reduction algorithm, t d The delay time set for the digital processing unit.

[0021] Preferably, two adjacent microphones and the speaker between them are used as a noise reduction unit, with the microphone closer to the noise source in the noise reduction unit serving as the reference microphone, and the other microphone in the noise reduction unit serving as the error microphone. The noise reduction algorithm includes:

[0022] The digital sound signal corresponding to the reference microphone is input into the FxLMS algorithm to obtain the noise reduction digital signal of the speaker in this noise reduction cycle;

[0023] An estimated noise reduction signal is obtained based on the digital audio signal corresponding to the reference microphone. The digital audio signal corresponding to the error microphone is input into the time series prediction algorithm to obtain the prediction error signal. The coefficients of the FIR filter in the FxLMS algorithm are updated using the estimated noise reduction signal and the prediction error signal to complete one noise reduction cycle for the noise reduction unit.

[0024] Preferably, the noise reduction algorithm is implemented independently in each noise reduction unit.

[0025] Preferably, the step of inputting the audio digital signal corresponding to the reference microphone into the FxLMS algorithm to obtain the noise-reduced digital signal of the speaker in this noise reduction cycle includes:

[0026] Let the coefficients and input of the FIR filter at time n be:

[0027] w(n)=[w1(n),w2(n),…,w L (n)] T

[0028] x(n)=[x c (n), x c (n-1), ..., x c (n-L+1)] T

[0029] In the formula, w(n) represents the coefficients of the FIR filter at time n, T is the transpose, L is the length of the FIR filter, and x(n) is the input of the FIR filter at time n. c (n) represents the digital audio signal corresponding to the reference microphone at time n;

[0030] The noise-reduced digital signal output by the FIR filter is:

[0031] y(n)=x(n) T w(n)

[0032] In the formula, y(n) is the denoised digital signal obtained at time n, and x(n) is the signal obtained at time n. T It is the transpose of x(n).

[0033] Preferably, the step of obtaining the estimated noise-reduced signal based on the digital sound signal corresponding to the reference microphone includes:

[0034] x′(n)=S′(z) T x(n)

[0035] In the formula, x′(n) is the estimated denoised signal obtained at time n, and S′(z) is the signal obtained at time n. T S′(z) is the transpose of S′(z), which is the secondary channel model, and the secondary channel model is an FIR filter.

[0036] Preferably, the secondary channel model modeling process is as follows: the digital processing unit plays white noise, the error microphone receives the white noise signal, and the secondary channel model is updated until the error between the digital signal corresponding to the white noise signal received by the error microphone and the signal calculated by the secondary channel model is less than a threshold, then the secondary channel model modeling is completed.

[0037] The secondary channel model update formula is as follows:

[0038] S′(z+1)=S′(z)+μ1wn(n)(wn(n)S′(z)-wn(n)S(z))

[0039] In the formula, S′(z+1) is the updated secondary channel model, S′(z) is the original secondary channel model, μ1 is the first step length value, wn(n) is the digital signal corresponding to the voltage of white noise, and wn(n)S(z) is the digital signal corresponding to the white noise signal received by the error microphone.

[0040] Preferably, updating the coefficients of the FIR filter in the FxLMS algorithm using the estimated denoised signal and the prediction error signal includes:

[0041] w(n+1)=w(n)+μ2x′(n)e′(n)

[0042] e′(n) = e′1(n) + e2′(n)

[0043] e′1(n)=ω e1 e(n)

[0044] e′2(n)=ω e2 e p (n)

[0045] In the formula, w(n+1) represents the coefficients of the updated FIR filter, w(n) represents the coefficients of the original FIR filter, μ2 represents the second step size, x′(n) represents the estimated denoised signal obtained at time n, e′(n) represents the prediction error signal obtained at time n, e′1(n) represents the first proportional error value, and ω e1 Let e(n) be the first weight value, e′2(n) be the error value corresponding to the digital audio signal of the microphone at time n, and ω be the second proportional error value. e2 e is the second weight value. p e(n) represents the prediction error value output by the time series prediction algorithm based on e(n).

[0046] The cascaded predictive active noise cancellation system provided by this invention has the following advantages compared with the prior art:

[0047] 1) A multi-stage noise reduction mode is adopted. In the event that the front-stage noise reduction unit fails due to excessive interference, the rear-stage unit compensates for its noise reduction function, thereby improving the noise reduction effect and enhancing the stability of noise reduction.

[0048] 2) In the multi-stage noise reduction mode, the error microphone data of the pre-stage is reused as the reference microphone data of the post-stage, which greatly reduces the amount of data, reduces the storage requirements of the digital processing unit, and lowers the cost of use and hardware.

[0049] 3) A weighted prediction-based error estimation method is adopted to improve the lag in updating the denoising filter coefficients caused by data misalignment in time. Furthermore, the true error ensures the validity of the input signal to the algorithm when sudden changes in the external environment cause abrupt changes in the error signal. Error estimation not only expands the range of the algorithm's update step size, improving the convergence speed of the denoising algorithm, but also makes the denoising error more accurate, thus improving the denoising effect.

[0050] 4) There is a settable delay time in the calculation of the prediction point. By setting this data properly, the error caused by the delay of circuit transmission and digital-to-analog / analog-to-digital conversion can be compensated, making the final prediction data more accurate. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of an active noise reduction method in the prior art;

[0052] Figure 2 This is a schematic diagram of the cascaded predictive active noise reduction system of the present invention;

[0053] Figure 3 This is a schematic diagram of the noise reduction unit of the present invention;

[0054] Figure 4 This is a schematic diagram of the microphone installation according to the present invention;

[0055] Figure 5 This is a schematic diagram of the noise reduction algorithm of the present invention;

[0056] Figure 6 This is a schematic diagram illustrating the calculation process of the FIR filter in this invention.

[0057] Figure 7 This is a schematic diagram of the secondary channel modeling process of the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0059] It should be noted that when a component is said to be "connected" to another component, it can be directly connected to the other component or it can be connected to a component in between; when a component is said to be "fixed" to another component, it can be directly fixed to the other component or it can be connected to a component in between.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0061] This embodiment proposes a scheme to reduce the noise of the pipeline using a cascaded / time series predictive FxLMS active noise reduction system.

[0062] like Figure 2 As shown, the noise reduction system proposed in this embodiment consists of the following parts: microphone (MIC), speaker (SPK), analog-to-digital converter, digital-to-analog converter and digital processing unit, wherein there are M speakers and M+1 microphones, where M is a positive integer greater than 1, and the upper limit is determined by the computing power of the digital processing unit.

[0063] The noise source is at one end of the pipe or inside the pipe, and the noise generated will be transmitted along the pipe to the other end, which is usually a place where people work or live.

[0064] Multiple microphones are spaced apart along the length of the pipe to collect noise signals within the pipe. A loudspeaker is placed between every two adjacent microphones, each loudspeaker connected to the pipe via a separate secondary channel. The loudspeakers play noise-reducing waves, which mix with the noise at a certain point in the pipe, producing a noise reduction effect and lowering the noise level beyond that point. An analog-to-digital converter (ADC) converts the analog signals collected by the microphones into digital signals, which are then transmitted to a digital processing unit (DMU). The DMU receives the digital audio signals from the ADC, executes a noise reduction algorithm based on the digital audio signals from two adjacent microphones, obtains the noise-reduced digital signal from the loudspeaker between the two microphones, and sends it to the digital-to-analog converter (DAC). The DAC receives the noise-reduced digital signal from the DMU, ​​converts it into a noise-reduced analog signal, and sends it to the loudspeaker, causing the loudspeaker to emit the corresponding noise-reduced wave for active noise reduction. Both the DAC and the analog-to-digital converter (ADC) include fixed-gain electrical signal amplifiers.

[0065] In this embodiment, two adjacent microphones and the speaker between them are considered as a single noise reduction unit. The microphone closer to the noise source within the noise reduction unit serves as the reference microphone, while the other microphone in the unit acts as the error microphone. In other words, the entire noise reduction system can be divided into N noise reduction cells, each containing two microphones and one speaker.

[0066] Since the noise reduction algorithm logic executed in different noise reduction units is the same, the only difference may be the noise reduction parameters. Therefore, for ease of description, this embodiment will use one noise reduction unit as an example for detailed explanation.

[0067] like Figure 3 The diagram shows an example of a noise reduction unit, the m-th noise reduction unit, where 1 ≤ m ≤ M. This noise reduction unit includes a speaker SPKm, which is connected to a conduit via a secondary channel of the m-th stage. For ease of description, the conduit is referred to as the main channel. Additionally, microphone MICm is the reference microphone, and microphone MICm+1 is the error microphone. Except for microphone MIC1, which serves only as the reference microphone, and microphone MICm+1, which serves only as the error microphone, all other microphones MICm serve as both the reference microphone for the m-th noise reduction unit and the error microphone for the (m-1)-th unit.

[0068] For any noise reduction unit m, it will satisfy the following requirements:

[0069] 1) Any microphone should be pointed in the direction of the noise source to avoid receiving reflected echoes.

[0070] 2) such as Figure 4 As shown, the microphone is placed at the center of the main channel cross-section.

[0071] 3) In the direction of the noise source, the angle ∠b between the secondary channel and the duct should be less than 90°. Theoretically, noise reduction can be achieved at any angle ∠b between 0° and 90°, but based on actual noise reduction effects, the smaller the angle ∠b, the better the noise reduction effect. Therefore, it is preferable to set the angle ∠b to be less than 15°. The actual angle can be adjusted according to the actual layout environment and process limitations.

[0072] 4) The sound-emitting part of the loudspeaker should be a centrally symmetrical regular shape and cover the entire m-th secondary channel without creating gaps.

[0073] 5) The microphone sampling rate is 48KHz, 96KHz, 144KHz or 192Khz.

[0074] 6) The mixing point of the noise wave emitted by the loudspeaker and the noise wave in the pipe on the central axis of the pipe is called the mixing point of the noise reduction unit. Then, the positions of the loudspeaker and the reference microphone in the same noise reduction unit satisfy the following relationship:

[0075]

[0076] In the formula, L micm L is the distance between the reference microphone MICm and the mixing point. spkm Let SPKm be the distance between the loudspeaker and the mixing point, and Vs be the speed of sound. Assume the speed of sound in both the primary and secondary channels is Vs. da The calibrated digital-to-analog conversion time for the digital-to-analog converter (this parameter is obtained through actual testing of different devices), t ad The calibrated analog-to-digital conversion time for the analog-to-digital converter (this parameter is obtained through actual testing of different devices), t p The time required for the digital processing unit to execute the noise reduction algorithm, t d The delay time set for the digital processing unit is used to match the equations.

[0077] In this embodiment, the digital processing unit executes a noise reduction algorithm to obtain the speaker output. The noise reduction algorithm executed can be the existing FxLMS algorithm or a custom noise reduction algorithm. To obtain a better noise reduction effect, this embodiment uses a noise reduction algorithm that combines the FxLMS algorithm with a time series prediction algorithm. This algorithm is valid for each noise reduction unit, but the parameters in each noise reduction unit are independent, such as... Figure 5 As shown, the noise reduction algorithm provided in this embodiment is as follows:

[0078] Step 1: Input the digital sound signal corresponding to the reference microphone into the FxLMS algorithm to obtain the digital noise reduction signal of the speaker in this noise reduction cycle.

[0079] In the figure, x(n) represents the signal sequence acquired by the microphone, y(n) represents the digital signal output to the speaker, y′(n) represents the noise reduction amplitude value reaching the error microphone after passing through the actual secondary channel, d(n) represents the noise amplitude value reaching the error microphone, and e(n) represents the digital signal corresponding to the signal acquired by the error microphone. S′(z) represents the modeled secondary channel model, and S(z) represents the actual secondary channel. x′(n) is the estimated noise reduction signal sequence obtained by estimating the modeled secondary channel model, and e′(n) is the prediction error signal after time series prediction. The above two sets of data are used in the LMS (Least Mean Square) algorithm to update the coefficients of the FIR filter.

[0080] In this embodiment, x(n) is a sequential queue with the same length as the FIR filter, and its update frequency is consistent with the microphone sampling rate. Each time the microphone sampling queue x(n) is updated, the last data item in the entire data queue is deleted, the entire queue is shifted one item to the right, and the newly sampled data is placed at the head of the queue. The noise reduction processing period is the sampling rate period; for example, for a sampling rate of 48kHz, the noise reduction processing period is 1s / 48kHz = 20.8us.

[0081] Figure 6 The diagram illustrates the calculation process of an FIR filter. An FIR filter consists of a set of data of length L, arranged in a specific order; these are called the FIR filter coefficients. The longer the coefficients of the FIR filter in the digital processing unit, the longer the processing time t for each noise-reduced signal. p The longer the length, the more it needs to satisfy the equation in point six of the requirements for each noise reduction unit. Generally, a length of 256 is chosen.

[0082] Let the coefficients and input of the FIR filter at time n be:

[0083] w(n)=[w1(n),w2(n),…,w L (n)] T

[0084] x(n)=[x c (n), x c (n-1), ..., x c (n-L+1)] T

[0085] In the formula, w(n) represents the coefficients of the FIR filter at time n, w1(n) represents the first-order parameters of the FIR filter, and w2(n) represents the second-order parameters of the FIR filter. L (n) represents the Lth-order parameter of the FIR filter, and the others can be understood similarly. r is the transpose, L is the length of the FIR filter, and x(n) is the input of the FIR filter at time n. c (n) represents the digital audio signal corresponding to the reference microphone at time n, x c (n-1) represents the digital audio signal corresponding to the reference microphone at time n-1, x c (n-L+1) represents the digital audio signal corresponding to the reference microphone at time n-L+1, and the others can be understood similarly. The digital audio signal corresponding to the reference microphone should be understood as the digital signal obtained by analog-to-digital conversion of the analog signal collected by the reference microphone.

[0086] The noise-reduced digital signal output by the FIR filter is:

[0087] y(n)=x(n) T w(n)

[0088] In the formula, y(n) is the denoised digital signal obtained at time n, and x(n) is the signal obtained at time n. T It is the transpose of x(n).

[0089] Step 2: Obtain the estimated noise reduction signal based on the digital sound signal corresponding to the reference microphone, input the digital sound signal corresponding to the error microphone into the time series prediction algorithm to obtain the prediction error signal, and use the estimated noise reduction signal and the prediction error signal to update the coefficients of the FIR filter in the FxLMS algorithm to complete one noise reduction cycle for the noise reduction unit.

[0090] The estimated noise-reduced signal is obtained based on the digital audio signal corresponding to the reference microphone, including:

[0091] x′(n)=S′(z) T x(n)

[0092] In the formula, x′(n) is the estimated denoised signal obtained at time n, and S′(z) is the signal obtained at time n. T S′(z) is the transpose of S′(z), which is the secondary channel model. The secondary channel model is an FIR filter with a length of a positive integer less than or equal to x(n) and greater than 1. Generally, the values ​​are of equal length.

[0093] like Figure 7 As shown, the secondary channel modeling process is as follows: The digital processing unit plays white noise, and the error microphone receives the white noise signal to update the secondary channel model until the error between the digital signal corresponding to the white noise signal received by the error microphone and the signal calculated by the secondary channel model is less than the threshold. Then the secondary channel modeling is completed.

[0094] The secondary channel model update formula is as follows:

[0095] S′(z+1)=S′(z)+μ1wn(n)(wn(n)S′(z)-wn(n)S(z))

[0096] In the formula, S′(z+1) is the updated secondary channel model, S′(z) is the original secondary channel model, μ1 is the first step length of the algorithm, wn(n) is the digital signal corresponding to the voltage of the white noise, and wn(n)S(z) is the digital signal corresponding to the white noise signal received by the error microphone. For each noise reduction unit, its secondary channel is independent, and each secondary channel needs to be modeled in a time-division manner to avoid front-end noise affecting the back-end modeling effect.

[0097] In each noise reduction cycle, the coefficients of the FIR filter need to be updated once. This update uses the predicted prediction error signal e′(n) and the estimated noise reduction signal x′(n). Therefore, updating the coefficients of the FIR filter in the FxLMS algorithm using the estimated noise reduction signal and the prediction error signal includes:

[0098] w(n+1)=w(n)+μ2x′(n)e′(n)

[0099] e′(n) = e′1(n) + e2′(n)

[0100] e′1(n)=ω e1 e(n)

[0101] e′2(n)=ω e2 e p (n)

[0102] In the formula, w(n+1) represents the coefficients of the updated FIR filter, w(n) represents the coefficients of the original FIR filter, μ2 represents the second step size of the algorithm, x′(n) represents the estimated denoised signal obtained at time n, e′(n) represents the prediction error signal obtained at time n, e′1(n) represents the first proportional error value, and ω e1 Let e(n) be the first weight value, e′2(n) be the error value corresponding to the digital audio signal of the microphone at time n, and ω be the second proportional error value. e2 e is the second weight value. p e(n) represents the prediction error value output by the time series prediction algorithm based on e(n).

[0103] The prediction error signal is obtained by weighting the predicted value of the acquired signal sequence after time series prediction with the current value, where the weighting coefficient ω is the weighting factor ω. e1 and ω e2 These are a set of empirical values ​​that need to be adjusted based on the specific pipeline. A typical value is ω. e1 =0.65 and ω e2 =0.35. Time series forecasting algorithms include, but are not limited to, exponential smoothing, ARIMA, and Holt linear trend method. The number of samples required for forecasting by a time series forecasting algorithm depends on the sampling rate. Assuming the sampling rate is mkHz (M = 48, 96, 144, or 192), and the historical sampling data sequence acquired by the microphone is e n = [e(n), e(n-1), ..., e(n-L+1)], this sequence is a queue, e(n) is the latest sampled value, which is the digital audio signal corresponding to the error microphone at time n, e(n-1) is the digital audio signal corresponding to the error microphone at time n-1, and so on, and arranged in the order of sampling forward), then the time consumed by each sampling point is The signal value to be predicted is the sampled signal from the time the noise reduction calculation is performed until the noise reduction result is applied to the microphone. Assuming the distance between these two points is s sampling points, then e p (n) = e(n+s) and

[0104] 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.

[0105] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A cascaded predictive active noise reduction system applied to duct noise reduction, characterized in that, The cascade prediction active noise reduction system comprises a microphone, a speaker, an analog-to-digital converter, a digital-to-analog converter and a digital processing unit, wherein: The microphone has a plurality of microphone, and the plurality of microphone is arranged in the pipeline in the length direction of the pipeline; The speaker has a plurality of speaker, and one speaker is arranged between two adjacent microphones, and each speaker is communicated with the pipeline through a separate secondary channel; The analog-to-digital converter is used for receiving an analog sound signal collected by the microphone, converting the analog sound signal into a digital sound signal and sending the digital sound signal to the digital processing unit; The digital processing unit is used for receiving the digital sound signal sent by the analog-to-digital converter, executing a noise reduction algorithm based on the digital sound signal corresponding to two adjacent microphones, obtaining a noise reduction digital signal of the speaker between the two adjacent microphones and sending the noise reduction digital signal to the digital-to-analog converter; The two adjacent microphones and the speaker between the two adjacent microphones are taken as a noise reduction unit, and a microphone closer to a noise source in the noise reduction unit is taken as a reference microphone, and another microphone in the noise reduction unit is taken as an error microphone, and the noise reduction algorithm comprises: The digital sound signal corresponding to the reference microphone is input into an FxLMS algorithm to obtain a noise reduction digital signal of the speaker in the current noise reduction period; An estimated noise reduction signal is obtained based on the digital sound signal corresponding to the reference microphone, a predicted error signal is obtained by inputting the digital sound signal corresponding to the error microphone into a time sequence prediction algorithm, and the coefficients of a FIR filter in the FxLMS algorithm are updated by using the estimated noise reduction signal and the predicted error signal, so that one noise reduction period for the noise reduction unit is completed; The digital-to-analog converter is used for receiving the noise reduction digital signal sent by the digital processing unit, converting the noise reduction digital signal into a noise reduction analog signal and sending the noise reduction analog signal to the speaker, so that the speaker emits a corresponding noise reduction sound wave to realize active noise reduction.

2. The cascaded predictive active noise reduction system of claim 1, wherein, The microphone is arranged at the center position of the pipeline section, and the microphone faces the noise source direction.

3. The cascaded predictive active noise reduction system of claim 1, wherein, The included angle between the secondary channel and the pipeline in the noise source direction is less than 90°.

4. The cascaded predictive active noise reduction system of claim 1, wherein, The mixing position of the noise reduction sound wave emitted by the speaker and the noise sound wave in the pipeline on the axis of the pipeline is called a mixing point, the two adjacent microphones and the speaker between the two adjacent microphones are taken as a noise reduction unit, and a microphone closer to the noise source in the noise reduction unit is taken as a reference microphone, and another microphone in the noise reduction unit is taken as an error microphone; The position of the speaker and the reference microphone in the same noise reduction unit satisfies the following relationship: ； wherein, is the distance between the reference microphone and the mixing point, is the distance between the loudspeaker and the mixing point, is the speed of sound, is the digital-to-analog conversion time for the digital-to-analog converter calibration, is the analog-to-digital conversion time for the analog-to-digital converter calibration, is the time required for the digital processing unit to execute the noise reduction algorithm, is the latency time set by the digital processing unit.

5. The cascaded predictive active noise reduction system of claim 1, wherein, The noise reduction algorithm is independently implemented in each noise reduction unit.

6. The cascaded predictive active noise reduction system of claim 1, wherein, The digital sound signal corresponding to the reference microphone is input into the FxLMS algorithm to obtain the noise reduction digital signal of the speaker in the current noise reduction period, and the noise reduction digital signal output by the FIR filter is: Let the coefficients of the FIR filter and the input be denoted by and x(n), respectively. ； ； wherein is the coefficient of the FIR filter at the time instant is the coefficient of the FIR filter at the time instant is the transpose, is the length of the FIR filter, is the coefficient of the FIR filter at the time instant is the input of the FIR filter at the time instant is the input of the FIR filter at the time instant is the digital signal corresponding to the sound of the reference microphone at the time instant The estimated noise reduction signal is obtained based on the digital sound signal corresponding to the reference microphone, and the noise reduction digital signal output by the FIR filter is: ； wherein is the first denotes the time instant at which the noise-reduced digital signal, is the transpose of 7. The cascaded predictive active noise reduction system of claim 6, wherein, ​ ； wherein is the first estimate noise reduced signal at time instant is the transpose of is a secondary path model, which is a FIR filter.​ 8. The cascaded predictive active noise reduction system of claim 7, wherein, The secondary path model modeling process is as follows: the digital processing unit plays white noise, the white noise signal received by the error microphone is used to update the secondary path model until the error between the digital signal corresponding to the white noise signal received by the error microphone and the signal calculated by the secondary path model is less than a threshold value, and then the secondary path model modeling is completed; The secondary path model update formula is as follows: ； In the formula, is the updated secondary path model, is the secondary path model before updating, is the first step value, is the digital signal corresponding to the voltage of white noise, is the digital signal corresponding to the white noise signal received by the error microphone.

9. The cascaded predictive active noise reduction system of Claim 1, wherein, The method for updating the coefficient of the FIR filter in the FxLMS algorithm by using the estimated noise reduction signal and the prediction error signal comprises the following steps: The method for updating the coefficient of the FIR filter in the FxLMS algorithm by using the estimated noise reduction signal and the prediction error signal comprises the following steps: ； ； ； ； In the formula, The coefficients of the updated FIR filter, The coefficients of the FIR filter before the update. This is the second step size value. For the first The estimated denoised signal obtained at time 1 For the first The prediction error signal obtained at time 10:00 First proportional error value, As the first weight value, No. The digital audio signal corresponding to the timing error microphone. This is the second proportional error value. As the second weight value, For time series prediction algorithms based on The output prediction error value.

Images