Digital modeling method, computer equipment and storage medium for nonlinear audio effects

Abstract

This application provides a digital modeling method, computer device, and storage medium for nonlinear audio effects. The method includes: acquiring sampled samples of an input signal from an audio source; sending the sampled samples of the input signal to the input of a target system and acquiring sampled samples of the output signal of the target system; based on a W-H nonlinear model, using the sampled samples of the input signal as the input of the W-H nonlinear model and the sampled samples of the output signal of the target system as the desired output of the W-H nonlinear model, and using adaptive moment estimation to calculate the parameters of the W-H nonlinear model; acquiring sampled samples of the desired output signal and sampled samples of the actual output signal of the W-H nonlinear model, performing a short-time Fourier transform on each sampled sample to obtain the amplitude spectrum of the two signals, calculating the difference between the two amplitude spectra, and fitting the difference using a cascaded filter to obtain the final W-H nonlinear model; and outputting the processed signal based on the final W-H nonlinear model.

Description

Technical Field

[0001] This invention belongs to the field of digital audio technology, and specifically relates to a digital modeling method, computer device and storage medium for nonlinear audio effects. Background Technology

[0002] Digital modeling of nonlinear effects has always been a subject of in-depth research in the field of digital audio effects. Various audio effects such as compression, exciter, overdrive, and distortion all fall under the category of nonlinear effects. However, analog devices capable of achieving these effects are typically expensive, bulky, and difficult to carry. Therefore, researching digital modeling methods for nonlinear effects is necessary.

[0003] Digital modeling of nonlinear effects primarily employs two approaches: white-box modeling and black-box modeling. White-box modeling, also known as forward modeling, utilizes all known information about the target system, including its circuit diagram, the specific parameters of electronic components, and their nonlinear characteristics. The advantage of white-box modeling is its ability to provide highly accurate modeling results. Its disadvantage lies in the requirement to obtain detailed features of the target system, and the complex topology and numerous nonlinear components of analog circuits can lead to excessively high computational costs for real-time processing. Black-box modeling, also known as inverse modeling, uses an abstract nonlinear system model to represent the target system and then calculates the modeling results using numerical methods. These abstract nonlinear models mainly include Volterra models, Chebyshev models, and artificial neural network models. The advantage of black-box modeling is that it does not require detailed features of the target system; as long as the target system conforms to the characteristics of the abstract nonlinear model, it can be modeled. Its disadvantage is that the modeling accuracy it can provide is relatively lower than that of white-box modeling.

[0004] For manufacturers whose core business is digital modeling of non-linear effects, white-box modeling, with its higher modeling accuracy, is undoubtedly the preferred method. However, for ordinary users, portability and flexibility are more important factors. Therefore, digital audio effects manufacturers typically choose white-box modeling as the core technology of their products, while offering black-box modeling as a unique feature for users to choose from. In existing technologies, non-linear models using black-box modeling are usually quite complex, with high computational costs and long processing times. This results in products equipped with black-box modeling technology generally being more expensive, and users experience longer waiting times when modeling target systems, leading to a poor user experience. Therefore, there is an urgent need for a modeling method with lower complexity and computational costs. Summary of the Invention

[0005] The main objective of this invention is to provide a digital modeling method for nonlinear audio effects, in order to solve the problems in the prior art, where nonlinear models using black-box modeling are usually quite complex, have high computational costs for model parameters, and require long computation times, resulting in generally high prices for products equipped with black-box modeling technology, and long waiting times and poor user experience when users are modeling target audio effects.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a digital modeling method for nonlinear audio effects, the method comprising:

[0008] Acquire sampled data of an audio source input signal, wherein the audio source includes any audio signal with wideband characteristics;

[0009] The sampled input signal is sent to the input of the target system, and the sampled output signal of the target system is acquired.

[0010] Based on the WH nonlinear model, the sampled samples of the input signal are used as the input of the WH nonlinear model, and the sampled samples of the target system output signal are used as the expected output of the WH nonlinear model. The parameters of the WH nonlinear model are calculated using adaptive moment estimation.

[0011] Obtain the sampled samples of the desired output signal and the sampled samples of the actual output signal of the WH nonlinear model. Perform a short-time Fourier transform on each sampled sample to obtain the amplitude spectrum of the two signals. Calculate the difference between the two amplitude spectra and fit the difference using a cascaded filter to obtain the final WH nonlinear model.

[0012] The processed signal is output based on the final WH nonlinear model.

[0013] Furthermore, the WH nonlinear model is composed of cascaded filter units and nonlinear units.

[0014] Furthermore, the specific parameters for calculating the WH nonlinear model include:

[0015] First, calculate the output of the FIR filter using the following formula:

[0016] y[n] = x[n] * h[n]

[0017] Where x[n] represents the input signal; y[n] represents the output signal; and h[n] represents the filter transfer function;

[0018] The output of the nonlinear unit is calculated using the following formula:

[0019]

[0020] Where, k p k n g p g n The parameters are used to adjust the response characteristics of the nonlinear curve, which is used to simulate the nonlinear characteristics of electronic components.

[0021] Then, calculate the output signal y of the target system. real [n] and the model's output signal y model The mean square error J of [n] in the time domain is shown in the following formula:

[0022]

[0023] Among them, y model [n] represents the model's output signal; y real [n] represents the output signal of the target system; N is the number of signal sample points;

[0024] Next, the partial derivatives needed in adaptive moment estimation are calculated, including...

[0025] Furthermore, it also includes updating the parameters of the WH model using adaptive moment estimation, with the calculation steps as follows:

[0026] (1) Calculate the moving average gradient v:

[0027]

[0028] (2) Calculate the moving average gradient v after error correction. corrected :

[0029]

[0030] (3) Calculate the moving average squared gradient s:

[0031]

[0032] (4) Calculate the error-corrected moving average squared gradient s corrected :

[0033]

[0034] (5) Update parameter θ:

[0035]

[0036] Where θ represents any parameter in the WH model; t represents the current iteration number of the model; β1 and β2 are used to determine the window size for calculating the moving average; and ε is used to prevent gradient explosion caused by an excessively small denominator.

[0037] Furthermore, it also includes using an adaptive learning rate decay method during parameter updates, specifically including:

[0038] (1) Calculate the slope of the mean square error J curve at the current iteration number;

[0039] (2) Set a threshold and determine whether the slope of the current iteration is less than the threshold. If yes, the model is considered to need to decay the learning rate and proceed to step (3). If no, the learning rate remains unchanged.

[0040] (3) Decrease the learning rate using the following formula:

[0041] α=α×r

[0042] Where α represents the learning rate and r represents the decay rate.

[0043] Furthermore, the signal processed based on the final WH nonlinear model output also includes:

[0044] The system allows users to participate in the listening test, providing them with input signal gain adjustment and adjusting the amount of gain based on user commands.

[0045] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method described in the first aspect above.

[0046] Thirdly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method described in the first aspect above.

[0047] The beneficial technical effects of the present invention are as follows:

[0048] The present invention provides a digital modeling method, computer device, and storage medium for nonlinear audio effects. It acquires an input signal, sends it to a target system, and then acquires the target system's output signal. Based on the WH model, it performs digital modeling of the target system using adaptive moment estimation. Furthermore, it introduces gain adjustment and adaptive learning rate decay to accelerate the model's convergence process. This addresses the problems in existing technologies where black-box modeling of nonlinear models is typically complex, computationally expensive, and time-consuming, leading to higher prices for products equipped with black-box modeling technology and longer user experience due to longer waiting times when modeling target audio effects. Attached Figure Description

[0049] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0050] Figure 1 This is a flowchart illustrating a digital modeling method for nonlinear audio effects according to an embodiment of the present invention.

[0051] Figure 2 This diagram illustrates how a user connects a device equipped with the method of this invention to other devices, using a Peavey 5150 tube speaker as the modeling object.

[0052] Figure 3 The waveforms are the actual output signals of the same electric guitar signal from the Peavey 5150 speaker and the output waveforms of the model obtained by the method of this invention.

[0053] Figure 4 This is a schematic diagram of the composition of a computer device provided in an embodiment of the present invention.

[0054] The accompanying drawings have illustrated specific embodiments of this disclosure, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concepts of this disclosure to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. The embodiments described below do not represent all embodiments consistent with this invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the claims.

[0056] In existing technologies, nonlinear models using black-box modeling are typically complex, with high computational costs for model parameters and long computation times. This results in generally higher prices for products equipped with black-box modeling technology, and users experience long waiting times and a poor user experience when modeling the target system. Therefore, there is an urgent need for a modeling method with lower complexity and computational costs.

[0057] To address the aforementioned problems, embodiments of the present invention provide a digital modeling method, computer device, and storage medium for nonlinear audio effects, thereby resolving the issues in the prior art.

[0058] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems will be described in detail below with reference to specific embodiments. The embodiments of the present invention will be described below with reference to the accompanying drawings.

[0059] Figure 1 A flowchart illustrating a digital modeling method for nonlinear audio effects provided in an embodiment of the present invention is shown below. Figure 1 As shown, the method includes:

[0060] S101. Obtain sampled data of the audio source input signal.

[0061] The audio source can be any audio signal with wide bandwidth characteristics, such as guitar playing signals, singer singing signals, and sine sweep signals, etc.

[0062] S102. Send the input signal to the input of the target system and obtain the sampled signal of the target system output signal;

[0063] S103. Based on the WH nonlinear model, the sampled samples of the input signal are used as the input of the WH nonlinear model, and the sampled samples of the target system output signal are used as the expected output of the WH nonlinear model. The parameters of the WH nonlinear model are calculated using adaptive moment estimation.

[0064] In this embodiment, the WH nonlinear model is composed of a cascaded FIR filter unit and a nonlinear unit. During the calculation of the model parameters, the output of the FIR filter is first calculated using the following formula:

[0065] y[n] = x[n] * h[n]

[0066] Where x[n] represents the input signal; y[n] represents the output signal; and h[n] represents the filter transfer function.

[0067] The output of the nonlinear unit is calculated using the following formula:

[0068]

[0069] Where, kp k n g p g n The parameters are used to adjust the response characteristics of the nonlinear curve, which is used to simulate the nonlinear characteristics of electronic components.

[0070] Then, calculate the output signal y of the target system. real [n] and the model's output signal y model The mean square error J of [n] in the time domain is shown in the following formula:

[0071]

[0072] Among them, y model [n] represents the model's output signal; y real [n] represents the output signal of the target system; N is the number of signal sample points.

[0073] Next, the partial derivatives needed in adaptive moment estimation are calculated, including...

[0074] The adaptive moment estimation is used to update the model parameters. The calculation steps are as follows:

[0075] (1) Calculate the moving average gradient v:

[0076]

[0077] (2) Calculate the moving average gradient v after error correction. corrected :

[0078]

[0079] (3) Calculate the moving average squared gradient s:

[0080]

[0081] (4) Calculate the error-corrected moving average squared gradient s corrected :

[0082]

[0083] (5) Update parameter θ:

[0084]

[0085] Where θ represents any parameter in the WH nonlinear model; t represents the current iteration number of the model; β1 and β2 are used to determine the window size for calculating the moving average; and ε is used to prevent gradient explosion caused by an excessively small denominator.

[0086] To accelerate model convergence and reduce computation time, the learning rate needs to be decayed during parameter updates. An adaptive learning rate decay method is applied here, and its steps are as follows:

[0087] (1) Calculate the slope of the mean square error J curve at the current iteration number;

[0088] (2) Set a threshold and determine whether the slope of the current iteration is less than the threshold. If yes, the model is considered to need to decay the learning rate, and proceed to step (3); otherwise, the learning rate remains unchanged.

[0089] (3) Decrease the learning rate using the following formula:

[0090] α=α×r

[0091] Where α represents the learning rate and r represents the decay rate.

[0092] S104. Obtain the sampled samples of the expected output signal and the actual output signal of the WH nonlinear model. Perform a short-time Fourier transform on each sampled sample to obtain the amplitude spectrum of the two signals. Calculate the difference between the two amplitude spectra and fit the difference using a cascaded filter to obtain the final WH nonlinear model.

[0093] Preferably, in this embodiment, the filter is an IIR filter.

[0094] S105. Output the processed signal based on the final WH nonlinear model.

[0095] At this point, the user is allowed to participate in the listening test, and the user is provided with the gain adjustment of the input signal, which is then determined by the user.

[0096] For example, using the Peavey 5150 tube amp as the modeling object, the user... Figure 2 The illustrated method involves connecting the device equipped with this technique to other equipment and inputting a short electric guitar signal (several seconds long). This signal is sent to a Peavey 5150 tube amplifier, which simultaneously returns its output signal. After several seconds, the modeling is complete, prompting the user to adjust the gain. Finally, the user confirms and saves the modeling data. Figure 3 The example shows the same electric guitar signal output from a Peavey 5150 amplifier and the model output obtained using this method. It can be seen that the model obtained using this method essentially reproduces the characteristics of the real Peavey 5150 amplifier.

[0097] Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention, such as... Figure 4As shown, the computer device 500 includes: a memory 501 and a processor 502;

[0098] Among them, memory 501 is used to store computer instructions that can be executed by the processor;

[0099] The processor 502 implements the various steps of the method in the above embodiments when executing computer instructions. For details, please refer to the relevant descriptions in the foregoing method embodiments.

[0100] Optionally, the memory 501 can be either independent or integrated with the processor 502. When the memory 501 is set up independently, the detection device also includes a bus for connecting the memory 501 and the processor 502.

[0101] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to perform the methods provided in the above embodiments.

[0102] This application also provides a computer program product, which includes: a computer program stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the solution provided in any of the above embodiments.

[0103] The above provides a detailed description of the digital modeling method, apparatus, device, and storage medium for nonlinear audio effects described in this application. The above examples are only for the purpose of helping to understand the method and core ideas of this application and are not intended to limit it. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principles of this application shall be considered equivalent substitutions and shall be included within the protection scope of this application.

Claims

1. A digital modeling method for nonlinear audio effects, characterized in that, The method includes: Acquire sampled data of an audio source input signal, wherein the audio source includes any audio signal with wideband characteristics; The sampled input signal is sent to the input of the target system, and the sampled output signal of the target system is acquired. Based on the WH nonlinear model, the sampled samples of the input signal are used as the input of the WH nonlinear model, and the sampled samples of the target system output signal are used as the expected output of the WH nonlinear model. The parameters of the WH nonlinear model are calculated using adaptive moment estimation. The WH nonlinear model is composed of cascaded filter units and nonlinear units. The parameters for calculating the WH nonlinear model specifically include: First, calculate the output of the FIR filter using the following formula: in, Indicates the input signal; Indicates the output signal; Represents the filter transfer function; The output of the nonlinear unit is calculated using the following formula: in, , , , The parameters are used to adjust the response characteristics of the nonlinear curve, which is used to simulate the nonlinear characteristics of electronic components. Then, calculate the output signal of the target system. With the model's output signal Mean square error in the time domain As shown in the following formula: in, This represents the output signal of the model; This represents the output signal of the target system; This represents the number of sampling points for the signal sample. Next, the partial derivatives needed in adaptive moment estimation are calculated, including... , , , , , , ; Obtain the sampled samples of the desired output signal and the sampled samples of the actual output signal of the WH nonlinear model. Perform a short-time Fourier transform on each sampled sample to obtain the amplitude spectrum of the two signals. Calculate the difference between the two amplitude spectra and fit the difference using a cascaded filter to obtain the final WH nonlinear model. The processed signal is output based on the final WH nonlinear model.

2. The digital modeling method for nonlinear audio effects as described in claim 1, characterized in that, It also includes updating the parameters of the WH model using adaptive moment estimation, with the calculation steps as follows: (1) Calculate the moving average gradient : (2) Calculate the moving average gradient after error correction : (3) Calculate the moving average squared gradient : (4) Calculate the error-corrected moving average squared gradient : (5) For parameters Update: in, Represents any parameter in the WH model; This indicates the current iteration number of the model; , The window size used to determine the moving average calculation; This is used to prevent gradient explosion caused by an excessively small denominator term; This represents the learning rate.

3. The digital modeling method for nonlinear audio effects as described in claim 2, characterized in that, This also includes using an adaptive learning rate decay method during parameter updates, specifically including: (1) Calculate the mean square error at the current iteration number. The magnitude of the slope of the curve; (2) Set a threshold and determine whether the slope of the current iteration is less than the threshold. If yes, the model is considered to need to decay the learning rate and proceed to step (3). If no, the learning rate remains unchanged. (3) Decrease the learning rate as follows: in, Indicates the learning rate. This indicates the attenuation rate.

4. The digital modeling method for nonlinear audio effects as described in any one of claims 1-3, characterized in that, The signal processed based on the final WH nonlinear model output also includes: The system allows users to participate in the listening test, providing them with input signal gain adjustment and adjusting the amount of gain based on user commands.

5. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method of any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as claimed in any one of claims 1-4.

Images