Imitated sound signal generation device, electronic musical instrument and non-linear system identification method

Abstract

To provide an imitated sound signal generation device in which imitation accuracy does not deteriorate even if a non-linear system with a large distortion rate is imitated.SOLUTION: A sound signal is inputted to a filter system including a first linear filter 10, a non-linear filter 30 and a second linear filter 20 and a simulated sound signal is generated. A primary component obtained by applying an inverse filter of a first input signal to an output signal from a non-linear system 200 when the first input signal having large amplitude is inputted is denoted as IRAH, and a primary component obtained by applying an inverse filter of a second input signal to an output signal from the non-linear system 200 when a random noise signal being a second input signal is inputted is denoted as IRAL. A characteristic of the first linear filter 10 is a time domain characteristic obtained by inversely-converting HAL / HAH and the characteristic 20 of the second linear filter is IRAH, when a transfer function corresponding to the primary component IRAL is denoted as HAL and a transfer function corresponding to the primary component IRAH as HAH.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to an imitative sound signal generating device, an electronic musical instrument, and a nonlinear system identification method. [Background technology]

[0002] For example, Patent Document 1 discloses a conventional technique for an imitation sound signal generating device that imitates a nonlinear system by placing linear filters before and after a nonlinear filter.

[0003] The imitation sound signal generating device of Patent Document 1 is a device that generates an imitation sound signal by inputting a sound signal into a filter system including a first linear filter, a nonlinear filter, and a second linear filter.

[0004] When an input signal (pink TSP signal) with a large amplitude is input to the above filter system, a distorted output signal is generated from the nonlinear filter. The first-order component obtained by applying the inverse filter of the input signal to the output signal from the nonlinear system is called IR. AH Let us assume that.

[0005] When an input signal (pink TSP signal) with an amplitude small enough that an undistorted output signal is obtained from the nonlinear filter is input to the above filter system, the first-order component obtained by applying the inverse filter of the input signal to the output signal from the nonlinear system is called IR. AL Let us assume that.

[0006] The characteristic of the first linear filter is the first-order component IR AL The transfer function corresponding to H AL The first-order component IR AH The transfer function H corresponding to AH Then, H AL / H AH The characteristic of the second linear filter is the first-order component IR AH It is.

[0007] In this way, in Patent Document 1, the characteristics of the first linear filter and the second linear filter can be obtained by inputting an input signal (pink TSP signal) with an amplitude that causes / does not cause distortion. [Prior art documents] [Patent documents]

[0008] [Patent Document 1] Patent No. 7072167 Summary of the Invention [Problem to be solved by the invention]

[0009] When a nonlinear system with a large distortion rate is imitated by the imitation sound signal generating device of Patent Document 1, the S / N ratio of the measurement result using an input signal with a small amplitude may deteriorate, resulting in a deterioration in imitation accuracy.

[0010] Therefore, an object of the present disclosure is to provide an imitation sound signal generating device in which imitation accuracy does not deteriorate even when imitating a nonlinear system with a large distortion rate. [Means for solving the problem]

[0011] According to the imitation sound signal generating device of the present disclosure, an imitation sound signal is generated by inputting a sound signal into a filter system in which a first linear filter, a nonlinear filter, and a second linear filter are connected in series in that order.

[0012] The first input signal is a signal having an amplitude that, when input to a nonlinear filter, results in a distorted output signal from the nonlinear filter, and whose frequency exponentially increases or decreases over time, and the second input signal is a random noise signal.

[0013] The first-order component obtained by applying the inverse filter of the first input signal to the output signal from the nonlinear system when the first input signal is input to the nonlinear system is called IR. AH Let us assume that.

[0014] The first-order component obtained by applying the inverse filter of the second input signal to the output signal from the nonlinear system when the second input signal is input to the nonlinear system is called IR. AL Let us assume that.

[0015] The characteristics of the first linear filter are the first-order component IR AL The transfer function corresponding to H AL Then, the first-order component IR AH The transfer function H corresponding to AH Then, H AL / H AH The time domain characteristic is obtained by inversely transforming the first-order component IR AH It is. Effect of the Invention

[0016] According to the imitation sound signal generating device of the present disclosure, imitation accuracy does not deteriorate even when imitating a nonlinear system with a large distortion rate. [Brief description of the drawings]

[0017] [Figure 1] The nonlinear system to be analyzed and the Wiener-Hammerstein model which is considered to be equivalent to it. [Diagram 2] An example of the characteristics of a nonlinear filter. [Diagram 3] (a) Frequency-amplitude characteristics of the Wiener model and the Hammerstein model when a measurement signal at a level that does not cause distortion is input to each of the Wiener model and the Hammerstein model. (b) Frequency-amplitude characteristics of the Wiener model and the Hammerstein model when a measurement signal at a level that causes distortion is input to each of the Wiener model and the Hammerstein model. [Figure 4] FIG. 1 is a diagram showing an example of a configuration for simulating a Wiener model. [Diagram 5]FIG. 11 is a diagram showing a comparison of the frequency-amplitude characteristics of the Wiener model and the first linear filter when a pink TSP signal at a level at which distortion occurs in the Wiener model is input and when a random noise signal at a level at which distortion occurs in the Wiener model is input. [Figure 6] An enlarged portion of Figure 5. [Figure 7] An example of the functional configuration of a nonlinear system identification device. [Figure 8] FIG. 1 is a flow diagram of a nonlinear system identification process. [Figure 9] 1 shows an example of the functional configuration of an imitation sound signal generating device. [Figure 10] Example of hardware configuration. [Figure 11] 13 shows the results (floor noise) of a nonlinear simulation performed for a method using a pink TSP signal and a method using a random noise signal. [Figure 12] A graph showing the results (linear characteristics) of a nonlinear simulation performed for a method using a pink TSP signal and a method using a random noise signal. [Figure 13] 13 shows an example of a functional configuration of a modified Wiener-Hammerstein model and an imitative sound signal generating device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] An embodiment of the present disclosure will be described with reference to the drawings.

[0019] In this embodiment, a system including a guitar amplifier and a microphone is taken as an example of the nonlinear system 200 to be analyzed (see FIG. 1). The guitar amplifier may be a stack amplifier or a combo amplifier. The amplifier may be a transistor or a vacuum tube. The term "comprises and comprises" means that the nonlinear system 200 strictly speaking includes not only the guitar amplifier and the microphone, but also the transfer characteristics of the space between the guitar amplifier and the microphone, the transfer characteristics of the signal in the wiring, and the like. Note that in the nonlinear system identification method of the present disclosure, there is no particular limitation on the nonlinear system to be analyzed. However, in the imitation sound signal generating device and electronic musical instrument of the present disclosure, the nonlinear system includes a device that outputs a sound signal.

[0020] First, in this embodiment, a model regarded as equivalent to a nonlinear system 200 to be analyzed is a Wiener-Hammerstein model 100 in which a first linear filter 10, a nonlinear filter 30, and a second linear filter 20 are connected in series in this order, as shown in FIG. 1. The first linear filter 10 is located on the input side of the Wiener-Hammerstein model 100, and the second linear filter 20 is located on the output side of the Wiener-Hammerstein model 100. The nonlinear filter 30 has a predetermined nonlinear characteristic. Therefore, in this embodiment, identifying a model regarded as equivalent to a nonlinear system to be analyzed is to specify the impulse responses of the first linear filter 10 and the second linear filter 20.

[0021] <Nonlinear filter> The nonlinear filter 30 preferably satisfies all of the conditions (1), (2), and (3) and has continuous nonlinear characteristics.

[0022] (1) When the amplitude of the input signal to the nonlinear filter 30 is in a continuous range around zero, including zero (hereinafter, the "continuous range around zero, including zero" is simply referred to as "near zero"), the amplitude of the output signal from the nonlinear filter 30 is expressed as a linear function of the amplitude of the input signal to the nonlinear filter 30, or is approximated by a linear function of the amplitude of the input signal to the nonlinear filter 30. However, the slope of the linear function is a positive value.

[0023] (2) Within the range of the amplitude of the input signal that can be input to the nonlinear filter 30 and within a range other than the vicinity of zero, there exists an amplitude of the input signal to the nonlinear filter 30 such that the amplitude of the output signal of the nonlinear filter 30 is smaller than the output of the linear function described above for the amplitude of the input signal to the nonlinear filter 30.

[0024] (3) It is time invariant.

[0025] Condition (1) is preferably rewritten as condition (1a).

[0026] (1a) When the amplitude of the input signal to nonlinear filter 30 is near zero, the amplitude of the output signal of nonlinear filter 30 is directly proportional to the amplitude of the input signal to nonlinear filter 30, or is approximately directly proportional to the amplitude of the input signal to nonlinear filter 30.

[0027] In the condition (1) or (1a), "near zero" may be rewritten as "a continuous range from zero to at least 5% of the maximum value or upper limit of the amplitude of the input signal that can be input to the nonlinear filter 30." "5%" is preferably rewritten as "10%," and more preferably as "20%."

[0028] Condition (2) is preferably rewritten as condition (2a).

[0029] (2a) Within the range of the amplitude of the input signal that can be input to the nonlinear filter 30, the amplitude of the output signal of the nonlinear filter 30 is expressed as an upwardly bounded continuous function.

[0030] Condition (2a) is preferably rewritten as condition (2b).

[0031] (2b) Within the range of the amplitude of the input signal that can be input to the nonlinear filter 30, the amplitude of the output signal from the nonlinear filter 30 is expressed as a monotonically increasing and upper bounded continuous function.

[0032] Condition (2b) is preferably rewritten as condition (2c).

[0033] (2c) Within the range of the amplitude of the input signal that can be input to the nonlinear filter 30, the amplitude of the output signal from the nonlinear filter 30 is expressed as a strictly monotonically increasing and upper bounded continuous function.

[0034] In each of the conditions (1), (1a), (2), (2a), (2b), and (2c), "amplitude" may be rephrased as "displacement relative to a predetermined reference value." However, when the term "displacement" is used, the conditions (2a), (2b), and (2c) can be rephrased as the conditions (2a'), (2b'), and (2c'), respectively.

[0035] (2a') Within the range of displacements of the input signal that can be input to the nonlinear filter 30, the displacement of the output signal of the nonlinear filter 30 is expressed as a bounded continuous function.

[0036] (2b') Within the range of displacements of the input signal that can be input to the nonlinear filter 30, the displacement of the output signal of the nonlinear filter 30 is expressed as a monotonically increasing and bounded continuous function.

[0037] (2c') In the range of the displacement of the input signal that can be input to the nonlinear filter 30, the nonlinear filter The displacement of the output signal of the filter 30 is expressed as a strictly monotonically increasing and bounded continuous function.

[0038] Furthermore, in each of the conditions (2a'), (2b'), and (2c'), the "continuous function" may preferably be rewritten as a "continuous odd function".

[0039] Preferably, the nonlinear filter 30 also satisfies condition (4).

[0040] (4) It has no frequency characteristics.

[0041] In identifying the nonlinear system 200 using the Wiener-Hammerstein model 100, the goal is to specify the impulse responses of the first linear filter 10 and the second linear filter 20, in other words, since the first linear filter 10 and the second linear filter 20 are responsible for the frequency characteristics of the nonlinear system 200, it is desirable for the nonlinear filter 30 to have as little frequency characteristics as possible.

[0042] More preferably, the nonlinear filter 30 also satisfies condition (5).

[0043] (5) When the amplitude of the input signal to nonlinear filter 30 is in a continuous range around the maximum value or upper limit, including the maximum value or upper limit, of the amplitude of the input signal that can be input to nonlinear filter 30 (hereinafter, the "continuous range around the maximum value or upper limit, including the maximum value or upper limit" will be simply referred to as "near the upper limit"), the amplitude of the output signal of nonlinear filter 30 is expressed as a constant function of the amplitude of the input signal to nonlinear filter 30, or is approximated by a constant function of the amplitude of the input signal to nonlinear filter 30. However, the value of the constant function is a positive value. Simply put, near the upper limit, the amplitude of the output signal of nonlinear filter 30 is almost constant, independent of the amplitude of the input signal to nonlinear filter 30.

[0044] In the condition (5), "near the upper limit" may be rewritten as "a continuous range from the maximum value or the upper limit to at least 95% of the maximum value or the upper limit of the amplitude of the input signal that can be input to the nonlinear filter 30." "95%" is preferably rewritten as "90%," and more preferably as "80%."

[0045] The nonlinear filter 30 has a characteristic h expressed by, for example, formula (1) or formula (2). nlIt has (x). x is the amplitude or displacement (however, when x is the amplitude, x ≥ 0). In Equation (2), each of A, B, C, α, and β is a predetermined positive definite constant.

Number

[0046] <Characteristics of Wiener-Hammerstein Model> In the Wiener-Hammerstein model 100 shown in FIG. 1, a model in which the first linear filter 10 and the non-linear filter 30 are connected in series in this order is known as a Wiener model. Also, in the Wiener-Hammerstein model 100, a model in which the non-linear filter 30 and the second linear filter 20 are connected in series in this order is known as a Hammerstein model.

[0047] In the Hammerstein model, regardless of the amplitude of the input signal input to the Hammerstein model, the characteristics of the second linear filter 20 remain in the amplitude of the output signal from the Hammerstein model.

[0048] In contrast, in the Wiener model, depending on the amplitude of the input signal input to the Wiener model, the characteristics of the first linear filter 10 may or may not remain in the amplitude of the output signal from the Wiener model. When a signal having an amplitude that can obtain an output signal that is not distorted from the non-linear filter 30 when input to the non-linear filter 30 is input to the Wiener model, the characteristics of the first linear filter 10 remain in the amplitude of the output signal from the Wiener model. However, generally, when a signal having an amplitude that can obtain a distorted output signal from the non-linear filter 30 when input to the non-linear filter 30 is input to the Wiener model, the characteristics of the first linear filter 10 do not remain in the amplitude of the output signal from the Wiener model.

[0049] FIG. 3 shows the characteristics of the first-order components (impulse responses) of the Wiener model and the Hammerstein model when a log-SS (logarithmic sine sweep) signal is used as the measurement signal, the nonlinear filter 30 has the nonlinear characteristics of formula (1), and the first linear filter 10 and the second linear filter 20 each have the characteristics of a peaking EQ (center frequency: 500 Hz, gain: -6 dB). FIG. 3(a) shows the results when the measurement signal is input to each model without amplification, and FIG. 3(b) shows the results when the measurement signal is amplified by 24 dB and input to each model. Note that the solid line and the dashed line overlap in FIG. 3(a). The above-mentioned difference between the Wiener model and the Hammerstein model can be understood from FIG. 3.

[0050] In this way, when a signal having an amplitude that generally results in a distorted output signal from the nonlinear filter 30 is input to the Wiener model, the characteristics of the first linear filter 10 do not remain in the amplitude of the output signal from the Wiener model. However, as an exception, if the input signal is a random noise signal (described later), a different result is obtained.

[0051] As shown in FIG. 4, when the parametric equalizer (PEQ) corresponding to the first linear filter 10 has a center frequency of 11,025 Hz, a gain of -20 dB, and a sharpness Q of 1.0, and the input signal is converted by the parametric equalizer (PEQ), amplified by 40 dB, and then clipped (corresponding to the process of obtaining a distorted output signal from the nonlinear filter 30), if the input signal is a pink TSP signal (Pink Time Stretched Pulse, synonymous with the log-SS signal described above), the characteristics of the first linear filter 10 (solid line graph in the same figure) disappear as shown in the dashed and dotted line graph in FIG. 5. On the other hand, if the input signal is a random noise signal, it can be seen that the characteristics of the first linear filter 10 remain almost the same as shown in the thin dashed line graph in the same figure. FIG. 6 is an enlarged view of the graph in FIG. 5 in the range of 10,000-12,000 Hz, and it can be seen that the characteristics of the first linear filter 10 are almost preserved for random noise signals.

[0052] In this simulation, the random noise signal was amplified to the extent that a distorted output signal was obtained from the nonlinear filter 30. However, the simulation results are not limited to this. It has been found that the characteristics of the first linear filter 10 are similarly preserved even when the random noise signal is input to the Wiener model with an amplitude that is not distorted (when the clipping in FIG. 4 is not performed).

[0053] <Random noise signal> It has been found that the input signal for which the characteristics of the first linear filter 10 are preserved in the Wiener model may be any irregular noise. Such irregular noise signals are called random noise signals. A typical example of a random noise signal is white noise. However, the condition for the input signal for which the characteristics of the first linear filter 10 are preserved is not necessarily that the power spectrum density is relatively uniform in intensity, as in white noise. For example, the characteristics of the first linear filter 10 are preserved even when a random noise signal with an uneven power spectrum density, such as colored noise, is used. Examples of colored noise include pink noise, Brownian noise, blue noise, purple noise, gray noise, and red noise. FVN (Frequency domain variant of Velvet Noise), which is a type of white noise, may be used as the random noise signal.

[0054] <Determining the impulse response of a linear filter> The following can be said from the above-mentioned characteristics of the nonlinear filter 30 and the respective characteristics of the Wiener model and the Hammerstein model.

[0055] The first input signal is a signal (e.g., a log-SS signal, a pink TSP signal) that has an amplitude (usually large; as a specific example, it is included in the “near upper limit” in condition (5)) that, when input to the nonlinear filter 30, results in a distorted output signal from the nonlinear filter, and whose frequency increases or decreases exponentially over time.

[0056] The second input signal is the random noise signal described above.

[0057] The transfer function H of the Wiener-Hammerstein model 100 shown in FIG. 1 when the second input signal (random noise signal) is input to the Wiener-Hammerstein model 100 ALis expressed by a transfer function H1H2, where H1 is the transfer function of the first linear filter 10 and H2 is the transfer function of the second linear filter 20. In this case, the second input signal (random noise signal) may have an amplitude that results in a distorted output signal from the nonlinear filter 30 when input to the nonlinear filter 30, or may have an amplitude that results in an undistorted output signal.

[0058] In addition, the transfer function H of the Wiener-Hammerstein model 100 when the first input signal is input to the Wiener-Hammerstein model 100 is AH is expressed by the transfer function H2 since the characteristics of the first linear filter 10 do not remain.

[0059] In this embodiment, the nonlinear system 200 to be analyzed is considered to be equivalent to the Wiener-Hammerstein model 100.

[0060] Therefore, the impulse response IR of the nonlinear system 200 when the first input signal is input to the nonlinear system 200 is AH is the impulse response IR2 of the second linear filter 20. That is, the impulse response IR AH The transfer function H obtained by Laplace transform or Z transform of AH is the transfer function H2 obtained by Laplace transform or Z transform of the impulse response IR2.

[0061] In addition, the impulse response IR of the nonlinear system 200 when the second input signal is input to the nonlinear system 200 is AL The transfer function H obtained by Laplace transform or Z transform of AL Transfer function H AH H divided by (=H2) AL / H AH is the transfer function H1 of the first linear filter 10. AL / H AH The time domain signal obtained by performing an inverse Laplace transform or an inverse Z transform on (=H1) is the impulse response IR1 of the first linear filter 10.

[0062] <Example> The nonlinear system identification process will be described with reference to Figs. 7 and 8. The signal generating unit 510 of the nonlinear system identification device 500 generates a signal whose frequency exponentially increases or decreases with time (step S1). An example of such a signal is a log-SS signal, which allows easy separation of harmonic distortion components in the time domain. The log-SS signal is also called a pink TSP signal, and is defined by, for example, equation (3) in the frequency domain. In equation (3), N is the signal length (N is an even number), j is the imaginary unit, J is the effective length (J is an integer), * is the complex conjugate, and log is the natural logarithm.

number

[0063] Then, the impulse response extraction unit 530 of the nonlinear system identification device 500 applies an inverse filter of the log-SS signal to the output signal from the nonlinear system 200, and extracts the first-order component (impulse response) IR AH (Step S3). The "inverse filter of the log-SS signal" is a filter that, when applied to the log-SS signal, produces an impulse signal. Since the log-SS signal is used, the result of applying the inverse filter of the log-SS signal to the output signal from the nonlinear system 200 is the first-order component IR AH and second-order and higher distortion components are separated, making it easy to detect the first-order component IR AH can be obtained.

[0064] Next, the signal generating unit 510 of the nonlinear system identification device 500 generates a random noise signal (step S4). The process of step S4 may be a conventionally known procedure for generating a general random noise signal.

[0065] Next, the random noise signal generated by the signal generating unit 510 is input to the nonlinear system 200 either amplified or unamplified by the control unit of the nonlinear system identification device 500 (step S5).

[0066] Then, the impulse response extraction unit 530 of the nonlinear system identification device 500 applies an inverse filter of the random noise signal to the output signal from the nonlinear system 200, and extracts the first-order component (impulse response) IR AL (Step S6).

[0067] 1st order component IR AH corresponds to the impulse response IR2 of the second linear filter 20.

[0068] In addition, the calculation unit 550 of the nonlinear system identification device 500 calculates the impulse response I AL The transfer function H corresponding to AL Impulse response IR AH The transfer function H corresponding to AH H divided by AL / H AH is subjected to an inverse Laplace transform or an inverse Z transform to obtain a time domain signal (step S7). This time domain signal corresponds to the impulse response IR1 of the first linear filter 10.

[0069] As a result, the model 100, which is deemed to be equivalent to the nonlinear system 200 being analyzed, is identified as a model in which a first linear filter 10 having an impulse response IR1, a nonlinear filter 30 having the above-mentioned predetermined nonlinear characteristics, and a second linear filter 20 having an impulse response IR2 are connected in series in that order.

[0070] The nonlinear system identification process may be performed in the order of step S4, step S5, step S6, step S1, step S2, step S3, and step S7, instead of the order shown in FIG.

[0071] When the nonlinear system identification device 500 is realized by a computer, for example, a dedicated machine configured with dedicated hardware or a general-purpose machine such as a personal computer, the nonlinear system identification device 500 includes an electronic circuit configured with components such as a processor 511, a memory 513, a hardware interface 515, and a bus 517 (see FIG. 10). The processor 511 executes arithmetic processing as necessary using data stored in a storage device such as the memory 513 according to a predetermined program, whereby the processor 511 realizes the functions of the signal generation unit 510, the impulse response extraction unit 530, and the calculation unit 550.

[0072] Next, the imitation sound signal generation process will be described with reference to Fig. 9. The imitation sound signal generation device 600 of this embodiment, which generates an imitation sound signal that imitates an original sound signal from a nonlinear system 200, includes an imitation sound signal generation unit 610 and a memory 613. The imitation sound signal generation unit 610 generates an imitation sound signal by inputting a sound signal to a time domain filter system in which a first linear filter 10 having an impulse response IR1, a nonlinear filter 30 having the above-mentioned predetermined nonlinear characteristics, and a second linear filter 20 having an impulse response IR2 are connected in series in this order. The first linear filter 10, which is a time domain filter, has a transfer function H AL / H AH The second linear filter 20, which is also a time domain filter, is made up of multipliers, adders and delay elements based on the transfer function H AHThe nonlinear filter 30 is realized as a processor that performs an operation according to a function such as equation (1) or equation (2), as a data table that associates an output value with an input value, or as a filter having the above-mentioned nonlinear characteristics. Information for configuring each of the first linear filter 10, the nonlinear filter 30, and the second linear filter 20 is stored in the memory 613.

[0073] It is more practical for the imitation sound signal generating device of the present disclosure to exist as a component of an electronic musical instrument rather than as an independent unit. In this case, the imitation sound signal generating device of the present disclosure may be a component of an electronic musical instrument that can be easily separated from the electronic musical instrument, or may be a one-sided evaluation of the electronic musical instrument focusing on a certain function of the electronic musical instrument itself. A typical example is an external effecter or an effecter built into the musical instrument. There is no particular limitation to the "electronic musical instrument including the imitation sound signal generating device" of the present disclosure, and this explanation should be taken into consideration when interpreting it. An example of an electronic musical instrument is a guitar amplifier, but this example is not limited to this example.

[0074] Of course, there is nothing to prevent the imitation sound signal generating device of the present disclosure from existing independently of an electronic musical instrument.

[0075] When the imitation sound signal generating device 600 is realized by a computer such as a dedicated machine configured with dedicated hardware or a general-purpose machine such as a personal computer, the imitation sound signal generating device 600 includes an electronic circuit configured with components such as a processor 611, a memory 613, a hardware interface 615, and a bus 617 (see FIG. 10). Note that, when the imitation sound signal generating device 600 is a component of an electronic musical instrument, these components may correspond to the processor, memory, hardware interface, bus, etc. of the electronic musical instrument. The processor 611 realizes the function of the imitation sound signal generating unit 610 by executing arithmetic processing using data stored in a storage device such as the memory 613 as necessary according to a predetermined program.

[0076] <Nonlinear simulation results for the conventional method and the method disclosed herein> 11 and 12 show the results of a nonlinear simulation (FIG. 11 shows floor noise, and FIG. 12 shows linear characteristics) performed in a conventional method for determining the characteristics of a first linear filter using a pink TSP signal and in the disclosed method for determining the characteristics of a first linear filter using a random noise signal. In the example shown in the figure, an FVN signal was used as the random noise signal. From the results shown in the figure, it can be seen that the floor noise of the linear characteristics is smaller when the FVN signal is used than when the pink TSP signal is used.

[0077] <Advantages of using random noise signals> In the conventional method using the pink TSP signal, it was necessary to input a signal with a small amplitude so that an undistorted output signal could be obtained from the nonlinear filter 30 in order to obtain the characteristics of the first linear filter. However, when trying to imitate a nonlinear system with a large distortion rate, the measurement result using the above-mentioned small amplitude signal had a poor S / N ratio, and the accuracy of the imitation tended to deteriorate. On the other hand, in the method of the present disclosure using the random noise signal, it is not necessary to input a signal with a small amplitude in order to obtain the characteristics of the first linear filter. For example, even if the amplitude of the random noise signal is increased to such an extent that a distorted output signal is obtained from the nonlinear filter 30, the characteristics of the first linear filter are preserved. As a result, even when trying to imitate a nonlinear system with a large distortion rate, the amplitude of the random noise signal can be freely adjusted, so that the S / N ratio of the measurement result does not deteriorate, and the accuracy of the imitation does not deteriorate.

[0078] <Modification> As a variation of the above embodiment, rather than using the nonlinear filter as the criterion for whether or not a nonlinear system to be imitated will be distorted, it is possible to easily generate an imitated sound signal by approximately applying the Wiener-Hammerstein model as the criterion for whether or not a nonlinear system to be imitated will be distorted.

[0079] 13, the imitation sound signal generating device 6000 of this modification is different from the above embodiment in that the first input signal has an amplitude that produces a distorted output signal from the "nonlinear system" to be imitated when the first input signal is input to the "nonlinear system" to be imitated, and the frequency of the signal increases or decreases exponentially with time, and the standard for whether the second input signal is distorted or not is the "nonlinear system." The other configurations are the same as those of the above embodiment.

[0080] Specifically, the imitation sound signal generating device 6000 of this modified example includes an imitation sound signal generating unit 6100 that generates an imitation sound signal by inputting a sound signal to a filter system (Wiener-Hammerstein model 1000) in which a first linear filter 100, a nonlinear filter 300, and a second linear filter 200 are connected in series in this order.

[0081] When the first input signal is input to the "nonlinear system" to be imitated, a distorted output signal is obtained from the "nonlinear system", and the signal has an amplitude and a frequency that exponentially increases or decreases over time.

[0082] The second input signal is a random noise signal.

[0083] The first-order component obtained by applying the inverse filter of the first input signal to the output signal from the nonlinear system when the first input signal is input to the nonlinear system is called IR. AH The first-order component obtained by applying the inverse filter of the second input signal to the output signal from the nonlinear system when the second input signal is input to the nonlinear system is defined as IR AL Then, the characteristic of the first linear filter 100 is the first-order component IR AL The transfer function corresponding to H AL Then, the first-order component IR AH The transfer function H corresponding to AH Then, H AL / H AH The characteristic of the second linear filter 200 is the time domain characteristic obtained by inversely transforming the first-order component IR AH It is.

[0084] The second input signal may be a random noise signal having an amplitude that, when input to the "nonlinear system" to be imitated, results in a distorted output signal from the "nonlinear system." As in the above embodiment, the random noise signal may be a white noise signal or a colored noise signal.

[0085] As in the above embodiment, the nonlinear filter may be characterized in that it satisfies all of the above (1)-(3) and has continuous nonlinear characteristics. Also, as in the above embodiment, the first input signal may be a log-SS signal.

[0086] Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to these embodiments. Various modifications and variations are permitted within the scope of the present disclosure. The selected and described embodiments are intended to illustrate the principles of the present disclosure and its practical application. The present disclosure is used in various embodiments with various modifications or variations, and the various modifications or variations are determined according to the expected applications. All such modifications and variations are intended to be included within the scope of the present disclosure as defined by the appended claims, and are intended to be accorded the same protection when interpreted according to the breadth that is fairly, legally and equitably afforded.

Claims

1. An imitation sound signal generating device that generates an imitation sound signal that imitates an original sound signal in a nonlinear system using an electronic circuit, an imitation sound signal generating unit that generates the imitation sound signal by inputting a sound signal into a filter system in which a first linear filter, a nonlinear filter, and a second linear filter are connected in series in this order; A first input signal is a signal having an amplitude such that a distorted output signal is obtained from the nonlinear filter when the first input signal is input to the nonlinear filter, and the frequency of the signal increases or decreases exponentially over time. A second input signal is a random noise signal. The first input signal is input to the nonlinear system, and a first-order component obtained by applying an inverse filter of the first input signal to the output signal from the nonlinear system is defined as an IR. ＡＨ The first-order component obtained by applying the inverse filter of the second input signal to the output signal from the nonlinear system when the second input signal is input to the nonlinear system is defined as IR ＡＬ Then, the characteristic of the first linear filter is the first-order component I ＡＬ The transfer function corresponding to ＡＬ and the first-order component IR ＡＨ The transfer function H ＡＨ Then, H ＡＬ / H ＡＨ The characteristic of the second linear filter is a time domain characteristic obtained by inversely transforming the first-order component I ＡＨ is Imitation sound signal generator.

2. The imitation sound signal generating device according to claim 1, The second input signal is has an amplitude that, when input to the nonlinear filter, results in a distorted output signal from the nonlinear filter Imitation sound signal generator.

3. The imitation sound signal generating device according to claim 1, The random noise signal is It can be a white noise signal or a colored noise signal. Imitation sound signal generator.

4. The imitation sound signal generating device according to claim 1, The nonlinear filter of the imitation sound signal generating device satisfies all of the following (1), (2), and (3) and has continuous nonlinear characteristics. (1) When the amplitude of the input signal to the nonlinear filter is in a continuous range around zero, including zero (hereinafter, the "continuous range around zero, including zero" is simply referred to as "near zero"), the amplitude of the output signal of the nonlinear filter is expressed as a linear function of the amplitude of the input signal to the nonlinear filter or is approximated by a linear function of the amplitude of the input signal to the nonlinear filter, where the slope of the linear function is a positive value. (2) Within the range of amplitudes of the input signal that can be input to the nonlinear filter and within a range other than the vicinity of zero, there exists an amplitude of the input signal to the nonlinear filter such that the amplitude of the output signal of the nonlinear filter is smaller than the output of the linear function for the amplitude of the input signal to the nonlinear filter. (3) It is time-invariant.

5. The imitation sound signal generating device according to claim 1, The first input signal is a log-SS signal. An imitation sound signal generating device comprising:

6. 6. An electronic musical instrument comprising the imitation sound signal generating device according to claim 1.

7. An imitation sound signal generating device that generates an imitation sound signal that imitates an original sound signal in a nonlinear system using an electronic circuit, an imitation sound signal generating unit that generates the imitation sound signal by inputting a sound signal into a filter system in which a first linear filter, a nonlinear filter, and a second linear filter are connected in series in this order; A first input signal is a signal having an amplitude such that a distorted output signal is obtained from the nonlinear system when the first input signal is input to the nonlinear system, and the frequency of the signal increases or decreases exponentially with time. A second input signal is a random noise signal. A first-order component obtained by applying an inverse filter of the first input signal to the output signal from the nonlinear system when the first input signal is input to the nonlinear system is defined as an IR filter. ＡＨ The first-order component obtained by applying the inverse filter of the second input signal to the output signal from the nonlinear system when the second input signal is input to the nonlinear system is defined as IR ＡＬ Then, the characteristic of the first linear filter is the first-order component I ＡＬ The transfer function corresponding to ＡＬ and the first-order component IR ＡＨ The transfer function H ＡＨ Then, H ＡＬ / H ＡＨ The characteristic of the second linear filter is a time domain characteristic obtained by inversely transforming the first-order component I ＡＨ is Imitation sound signal generator.

8. The imitation sound signal generating device according to claim 7, The second input signal is has an amplitude that, when input to said nonlinear system, results in a distorted output signal from said nonlinear system Imitation sound signal generator.

9. The imitation sound signal generating device according to claim 7, The random noise signal is It can be a white noise signal or a colored noise signal. Imitation sound signal generator.

10. A method for identifying a model that is considered equivalent to a nonlinear system to be analyzed, comprising the steps of: the model is a model in which a first linear filter, a nonlinear filter, and a second linear filter are connected in series in this order; a first input signal having an amplitude that, when input to the nonlinear filter, results in a distorted output signal from the nonlinear filter, and a frequency that exponentially increases or decreases over time; The second input signal is a random noise signal; A first-order component IR is obtained by applying an inverse filter of the first input signal to an output signal from the nonlinear system when the first input signal is input to the nonlinear system. ＡＨ and A first-order component IR is obtained by applying an inverse filter of the second input signal to an output signal from the nonlinear system when the second input signal is input to the nonlinear system. ＡＬ and having The above primary component IR ＡＬ The transfer function corresponding to ＡＬ and the first-order component IR ＡＨ The transfer function H ＡＨ Then, H ＡＬ / H ＡＨ The time domain characteristic obtained by inversely transforming the above is defined as the characteristic of the first linear filter, The above primary component IR ＡＨ is the characteristic of the second linear filter. Nonlinear system identification methods.

11. A method for identifying a model that is considered equivalent to a nonlinear system to be analyzed, comprising the steps of: the model is a model in which a first linear filter, a nonlinear filter, and a second linear filter are connected in series in this order; a first input signal having an amplitude that, when input to the nonlinear system, results in a distorted output signal from the nonlinear system, and a frequency that exponentially increases or decreases over time; The second input signal is a random noise signal; A first-order component IR is obtained by applying an inverse filter of the first input signal to an output signal from the nonlinear system when the first input signal is input to the nonlinear system. ＡＨ and A first-order component IR is obtained by applying an inverse filter of the second input signal to an output signal from the nonlinear system when the second input signal is input to the nonlinear system. ＡＬ and having The above primary component IR ＡＬ The transfer function corresponding to ＡＬ and the first-order component IR ＡＨ The transfer function H ＡＨ Then, H ＡＬ / H ＡＨ The time domain characteristic obtained by inversely transforming the above is defined as the characteristic of the first linear filter, The above primary component IR ＡＨ is the characteristic of the second linear filter. Nonlinear system identification methods.

Images