JP6304643B2 - Nonlinear distortion reduction apparatus, method, and program for speaker - Google Patents

Description

  The present invention relates to a speaker non-linear distortion reduction apparatus, method, and program.

  The speaker is expected to faithfully convert the input signal into sound. However, since the speaker is distorted and causes deterioration in sound quality, it is desired to reduce the distortion.

  Speaker distortion includes linear distortion and nonlinear distortion. Linear distortion can be modeled as a frequency characteristic (linear function) of a speaker in the frequency domain. Therefore, it can be reduced relatively easily by calculating the inverse filter from the inverse characteristic of the frequency characteristic and convolving it with the sound source.

  Since nonlinear distortion cannot be modeled as a linear function, it is difficult to obtain the inverse characteristic of nonlinear distortion. As a method for reducing this nonlinear distortion, a Volterra filter (see, for example, Non-Patent Document 2) and a mirror filter (see, for example, Non-Patent Document 5) have been proposed.

  However, since the Volterra filter has a large amount of calculation, it is difficult to reduce high-order nonlinear distortion. Also, mirror filters generally have a lower amount of reduction of nonlinear distortion than Volterra filters due to modeling errors and the like.

  Hereinafter, a method for reducing nonlinear distortion using a Volterra filter will be described.

A system with nonlinear inputs and outputs can be expressed using Volterra series expansion. If the Volterra kernel is finite, the Volterra series expansion can be expressed as:

Here, x (n) and y (n) are the sampled input signal and output signal, h ₁ (k ₁ ), h ₂ (k ₁ , k ₂ ), h ₃ (k ₁ , k ₂ ), respectively. , K ₃ ),... Are 1, 2, 3,... Discrete Volterra kernels, and N is the system length of the filter.

The transfer function of the nonlinear distortion of the speaker can be modeled as a Volterra kernel. The Volterra kernel is obtained by using a frequency response method or an adaptive Volterra filter, and an inverse system is constructed. The calculation amount of this method is O (N ⁿ ) when nonlinear distortion up to the nth order is targeted from the number of coefficients to be obtained. For this reason, it is difficult to reduce high-order nonlinear distortion, and practically, nonlinear distortion up to the third order is targeted.

  Next, a method for reducing nonlinear distortion using a mirror filter will be described.

  The mirror filter models the speaker based on the nonlinear operating equation of the speaker system. The equation of motion has parameters for each of linear and non-linear, and each parameter is obtained from the characteristics of impedance, diaphragm displacement, etc., and a correction filter is generated from the obtained parameters.

  The amount of calculation required for generation differs depending on the method of realizing the mirror filter, and is small in the method realized by the secondary IIR filter (for example, see Non-Patent Document 6). However, since the mirror filter does not completely describe the actual operation of the speaker, a modeling error occurs. Therefore, there is a limit to the reduction of nonlinear distortion (see, for example, Non-Patent Document 2).

JP 2005-80314 A Japanese Patent Laid-Open No. 11-55782

Akira Sugawara and Akira Kiryu, “Increase in Nonlinear Distortion of Speaker with Expansion of Frequency Band”, Acoustical Society Journal, 56, 549-555 (2000). Yoshinobu Kajikawa, “Nonlinear distortion correction of speaker system by signal processing technology”, Journal of Acoustical Society, 67, 470-475 (2011). Yoshinobu Kajikawa, “What is Volterra filter?-Its application to acoustic system”, Journal of Acoustical Society, 60, 273-277 (2004). V.J.Mathews, “Adaptive polynomial filters”, IEEE Signal Process. Mag., 8, 10-26 (1991). Rika Nakao, Yoshinobu Kajikawa, Yasuo Nomura, “Examination of nonlinear distortion correction method for speaker system using Mirror filter”, IEICE Technical Report, EA206-101, 21-26 (2007). Kazumasa Takemura, Tetsuro Nakatake, Joji Kasai, “Reduction of nonlinear distortion of electrodynamic loudspeakers by nonlinear second-order IIR filter”, IEICE Technical Report, EA96-74, 65-72 (1996). Hiroshi Nakajima, “Sidelobe Minimum Beamforming Based on Minimax Code”, Journal of the Acoustical Society of Japan, 63, 341-352 (2007).

  As described above, the method of reducing nonlinear distortion using a Volterra filter or a mirror filter requires a large amount of calculation and a large amount of parameters to be estimated. There was a problem.

  It is an object of the present invention to provide a speaker non-linear distortion reduction apparatus, method, and program capable of reducing speaker non-linear distortion with a small amount of calculation without modeling the system strictly with a non-linear function. .

In order to solve the above-mentioned problem, a non-linear distortion reducing apparatus for a speaker according to claim 1 is a measurement means for measuring an impulse response from a speaker to a microphone, a sound source signal output from the speaker, and the sound source signal as the speaker. A calculation means for calculating a non-linear component of the output signal based on an output signal from the speaker input from the microphone and input from the microphone, and a measurement result of the impulse response; and a measurement result of the non-linear component and the impulse response And a correction means for directly correcting the sound source signal.

  According to the present invention, it is possible to reduce the nonlinear distortion of the speaker with a small amount of calculation without strictly modeling the system with a nonlinear function.

  According to a second aspect of the present invention, the calculating means outputs a corrected sound source signal obtained by correcting the sound source signal by the correcting means, and the corrected sound source signal output from the speaker and input from the speaker. Based on the measurement result of the correction output signal and the impulse response, the processing for calculating the nonlinear component of the correction output signal, and the correction means based on the measurement result of the nonlinear component of the correction output signal and the impulse response The process of correcting the corrected sound source signal again may be repeated until a predetermined end condition is satisfied.

In addition, as described in claim 3, the calculation unit calculates the nonlinear component for each frequency band, and the correction unit calculates this time out of the nonlinear components calculated for each frequency band. For the frequency band in which the nonlinear component is equal to or greater than the previously calculated nonlinear component, the sound source signal may be corrected with the nonlinear component calculated this time as zero.
According to a fourth aspect of the present invention, the calculation means calculates a linear component that is an ideal frequency component of the output signal based on a value obtained by Fourier transforming the sound source signal, the output signal, and the impulse response. And calculating a nonlinear component of the output signal based on a Fourier transform of the linear component and the output signal, and the correcting means calculates a value obtained by dividing the nonlinear component by a value obtained by Fourier transforming the impulse response. The sound source signal may be directly corrected by subtracting from the sound source signal.

The method for reducing nonlinear distortion of a speaker according to claim 5 includes the step of measuring an impulse response from the speaker to the microphone, a sound source signal output from the speaker, and the sound source signal output from the speaker and input by the microphone. Calculating a nonlinear component of the output signal based on the output signal from the speaker and the measurement result of the impulse response; and directly calculating the sound source signal based on the measurement result of the nonlinear component and the impulse response. And a step of correcting.

The non-linear distortion reduction program for a speaker according to claim 6 causes a computer to function as the non-linear distortion reduction apparatus for a speaker according to any one of claims 1 to 4 .

  According to the present invention, it is possible to reduce the nonlinear distortion of a speaker with a small amount of calculation without strictly modeling the system with a nonlinear function.

It is a block diagram of the nonlinear distortion reduction apparatus of a speaker. It is a block diagram of a process until the sound source signal after correction | amendment is obtained from a sound source signal. It is a flowchart of a nonlinear distortion reduction process. It is a graph which shows the frequency characteristic of an input signal (sound source signal). It is a graph which shows the frequency characteristic of the output signal of the speaker with respect to an input signal. It is a graph which shows the frequency characteristic of the input signal correct | amended (1st time). It is a graph which shows the frequency characteristic of the output signal of a speaker with respect to the input signal correct | amended (1st time). It is a graph which shows the frequency characteristic of the input signal correct | amended (2nd time). It is a graph which shows the frequency characteristic of the output signal of a speaker with respect to the input signal correct | amended (2nd time). It is a graph which shows the reduction level of the nonlinear distortion with respect to the frequency | count of correction | amendment. It is a graph which shows the reduction level of the nonlinear distortion of various sound source signals in an anechoic room and a general room.

  Hereinafter, embodiments of the present invention will be described. In the present embodiment, a case where the present invention is applied to a speaker system as a nonlinear system will be described.

(Nonlinear system model)
First, the nonlinear system model of the present invention will be described. The output signal y (n) of the nonlinear system can be modeled by the following equation with the input signal as x (n).

ここで、Ｘは関数Ｘ（ω）を引数とすることを表し、Ｅ（ω、Ｘ）は入力の周波数特性Ｘにより変わる非線形歪の周波数特性を表す。ここで、Ｅ（ω、Ｘ）は、Ｘに対してなめらかな連続関数としてモデル化できると仮定する。 Here, h (n) is the impulse response (linear component) of the system, N is the length of the impulse response, * in the equation represents the convolution operator, and x represents the function x (n) as an argument. , E (n, x) represents a nonlinear distortion signal that varies depending on the input x. Moreover, when both sides are Fourier-transformed and converted to the frequency domain, the following equation is obtained.

Here, X represents that the function X (ω) is used as an argument, and E (ω, X) represents the frequency characteristic of nonlinear distortion that varies depending on the input frequency characteristic X. Here, it is assumed that E (ω, X) can be modeled as a smooth continuous function with respect to X.

(Principle of the present invention)
Next, the principle of the present invention will be described. First, the nonlinear component E (ω, X) is calculated by the following equation.

Next, the corrected input signal is obtained by the following equation.

ここで、

である。線形出力Ｘ（ω）・Ｈ（ω）に比べて非線形歪Ｅ（ω、Ｘ）の大きさが十分小さい（｜ＸＨ｜≫｜Ｅ｜）と仮定すると、

が成り立つ。そのため非線形歪である上記（７）式の右辺の第１項と第２項とは近い値となり、

となる。すなわち、理論上非線形歪が低減する。 The output signal when the corrected input signal obtained by the above equation (5) is input to the system H (ω) is expressed by the following equation.

here,

It is. Assuming that the magnitude of the nonlinear distortion E (ω, X) is sufficiently small (| XH | >> | E |) compared to the linear output X (ω) · H (ω),

Holds. Therefore, the first term and the second term on the right side of the equation (7), which is nonlinear distortion, are close to each other.

It becomes. That is, theoretically nonlinear distortion is reduced.

(Recursive processing)
Next, recursive processing will be described. In order to further reduce the nonlinear component, recursive processing is performed in the present embodiment. The recursive processing is used for solving a nonlinear problem (for example, see Non-Patent Document 7). In the following description, the argument ω is omitted for simplification of description.

The second input signal is obtained by the following equation so as to reduce the distortion after the first correction obtained by the above equation (7).

The output signal corresponding to the second input signal is expressed by the following equation.

  Similarly, the third input signal is obtained so as to reduce the distortion after the second correction, that is, the second term on the right side of the equation (11).

ここで、Ｘ_−１＝０、Ｅ（０）＝０と定義する。以下、再帰的補正により非線形歪が理論上低減することを示す。 If the first, second, and third input signals are X ₀ , X ₁ , and X ₂ , and the _first , _second , and third output signals are Y ₀ , Y ₁ , and Y ₂ , respectively, the number of corrections is L ( L = 1, 2,...

Here, X ₋₁ = 0 and E (0) = 0 are defined. Hereinafter, it will be shown that the nonlinear distortion is theoretically reduced by recursive correction.

In the above (principle of the present invention), it was shown that | D (X _L , X _L−1 ) | << | D (X _L−1 , X _L−2 ) | at L = 1. Assuming that | D (X _k , X _k−1 ) | << | D (X _k−1 , X _k−2 ) | holds true when L = k, from the above equation (12), | X _{k + 1} −X _k | << | X _k −X _k−1 |, and the update amount of X decreases. For this reason, | D (X _{k + 1} ), X _k ) | << | D (X _k , X _k−1 ) |, and it can be seen that L = k + 1 holds.

In FIG. 2, illustrating a process to the sound source signal X _{L + 1} corrected from the sound source signal X _L is obtained in block diagram.

  (Selection of correction frequency band)

According to the preliminary experiments by the present inventors, when a nonlinear distortion component having a low level due to background noise is fed back to the input signal and corrected, a phenomenon in which the distortion increases is seen. In order to prevent this, in the present embodiment, correction frequency band selection processing is performed in which the correction amount of the frequency component whose distortion component is increased by correcting the nonlinear distortion is set to zero. The nonlinear distortion after selection is calculated by the following equation.

(14) is calculated again X _L from the nonlinear distortion calculated by the equation, again calculated nonlinear distortion from the response after output from the speaker by the equation (14).

(Device configuration)
The apparatus configuration of the speaker non-linear distortion reducing apparatus will be described below. FIG. 1 shows the configuration of a speaker non-linear distortion reduction apparatus 10.

  As shown in FIG. 1, the speaker nonlinear distortion reduction apparatus 10 includes a controller 12, a speaker 14, a microphone 16, and an AD-DA converter 18.

  The controller 12 includes a CPU (Central Processing Unit) 12A, a ROM (Read Only Memory) 12B, a RAM (Random Access Memory) 12C, a non-volatile memory 12D, and an input / output interface (I / O) 12E via a bus 12F. It is a connected configuration. An AD-DA converter 18 is connected to the I / O 12E.

  The AD-DA converter 18 converts the digital audio signal output from the controller 12 into an analog audio signal and outputs the analog audio signal to the speaker 14, and converts the analog audio signal input from the microphone 16 into a digital audio signal. 12 is output.

  The controller 12 executes nonlinear distortion reduction processing that will be described later. In this embodiment, the control program for the nonlinear distortion reduction process is stored in advance in the nonvolatile memory 12D as an example, and is executed by the CPU 12A reading the control program stored in advance. Alternatively, the control program may be recorded on a recording medium such as a CD-ROM and executed by reading it with a CD-ROM drive or the like.

  (Nonlinear distortion reduction processing)

  Hereinafter, the nonlinear distortion reduction process executed by the CPU 12A of the controller 12 will be described with reference to the flowchart shown in FIG.

  First, in step 100, an impulse response h (n) from the speaker 14 to the microphone 16 is measured. In the present embodiment, as an example, an impulse response is measured using a TSP (Time-Stretched Pulse) method, and the measurement result is stored in the nonvolatile memory 12D.

  In step 102, the sound source signal x (n) is read from the audio file stored in the nonvolatile memory 12D.

  In step 104, the sound source signal x (n) is output to the AD-DA converter 18, converted into a digital audio signal, and output from the speaker 14. As a result, the signal output from the speaker 14 is input to the microphone 16, and the analog audio signal input to the microphone 16 is converted into a digital audio signal by the AD-DA converter 18 and output to the controller 12. The input digital audio signal is stored in the nonvolatile memory 12D as the output signal y (n).

In step 106, the sound source signal x (n), the output signal y (n), and the impulse response h (n) are Fourier-transformed to obtain X ₀ , Y ₀ , and H, and a linear component that is an ideal frequency component of the output signal. X ₀ H is calculated.

  In step 108, the number of corrections L is initialized (L = 0).

In step 110, based on X ₀ H and Y _L obtained in step 106, a nonlinear component D (X _L , X _L-1 ) is obtained from equation (13).

In step 112, the correction signal X _{L + 1} = X _L −D (X _L , X _L−1 ) / H is obtained from equation (12).

In step 114, X _{L + 1} obtained in step 112 is subjected to inverse Fourier transform to obtain a corrected output signal x _{L + 1} (n), which is output from the speaker 14 via the AD-DA converter 18, and y _{L + 1} (n). Is stored in the non-volatile memory 12D, and y _{L + 1} is obtained by Fourier transform. Further, a nonlinear component D (X _{L + 1} , X _L ) for X _{L + 1 is obtained} from the linear component X ₀ H.

In step 116, calculation according to equation (14) is executed as correction frequency selection processing. That is, for the frequency component satisfying | D (X _L , X _L-1 ) |> | D (X _{L + 1} , X _L ) |, the left side of the equation (14) is D (X _L , X _L-1 ), For the frequency component satisfying | D (X _L , X _L−1 ) | ≦ | D (X _{L + 1} , X _L ) |, the left side of the equation (14) is set to 0.

In step 118, D (X _L , X _L-1 ) in equation (12) is replaced with the value on the left side of equation (14) obtained in step 116, and then X _{L + 1} is obtained by equation (12).

を求める。 In step 120, the same processing as in step 114 is performed on XL _{+ 1} obtained in step 118 to obtain a nonlinear component.

Ask for.

ここで、Ｐ（Ｄ_Ｌ）は補正なしの出力信号Ｙ_０のレベルを基準とした全周波数帯域での非線形歪のレベルＤ_Ｌ＝Ｄ（ω、Ｘ_Ｌ、Ｘ_Ｌ−１）を表し、次式で計算される。すなわち、Ｐ（Ｄ_Ｌ）は、出力信号全体のパワーに対する非線形歪のパワーの割合を示す。なお、単位はｄＢである。

ここで、ω_ｍは離散角周波数、ｍ（ｍ＝１、２、…）は周波数インデックス、ＮＦＦＴはＦＦＴ長、ｄはアルゴリズム終了基準値（デフォルト値：１）である。また、システムでの窓は矩形窓、サンプリング周波数は４４１００Ｈｚ、フレーム長はＦＦＴ長と同じであり、ＮＦＦＴ＝２^ａで計算する。なお、ａ＝ｃｅｉｌ（ｌｏｇ_２ｓ）であり、ｃｅｉｌ（ｎ）は、ｎ以上である最小の整数を出力する関数であり、ｓは信号長を表す。 In step 122, it is determined whether or not the following equation is satisfied.

Here, P (D _L ) represents the level of nonlinear distortion D _L = D (ω, X _L , X _L−1 ) in the entire frequency band with reference to the level of the output signal Y ₀ without correction. Calculated by the formula. That is, P (D _L ) indicates the ratio of the power of nonlinear distortion to the power of the entire output signal. The unit is dB.

Here, ω _m is a discrete angular frequency, m (m = 1, 2,...) Is a frequency index, NFFT is an FFT length, and d is an algorithm end reference value (default value: 1). The windows in the system is a rectangular window, the sampling frequency is 44100Hz, the frame length is the same as the FFT length is calculated with NFFT = ^{2 a.} Note that a = ceil (log ₂ s), ceil (n) is a function that outputs the smallest integer that is greater than or equal to n, and s represents the signal length.

  If the above equation (16) is satisfied, that is, if the power of the nonlinear distortion is large, the process proceeds to step 124, and if not, the process proceeds to step 126.

In step 124, D (X _{L + 1} , X _L ) before correction is set as a nonlinear component at the next correction.

On the other hand, in step 126, the corrected nonlinear component (15) is set as a nonlinear component at the time of the next correction, and the nonlinear component (15) is overwritten on D (X _{L + 1} , X _L ).

In step 128, it is determined whether or not the following equation is satisfied.

Here, the first term on the right side of the above equation (18) represents the minimum value of P (D (X _k , X _k−1 )) in k = 1 to L. That is, in step 128, the value of the nonlinear distortion power calculated this time is a value obtained by adding d (1 dB as an example in the present embodiment) to the minimum nonlinear distortion power among the nonlinear distortion powers calculated in the correction so far. It is judged whether it is larger than.

  If the above equation (18) is satisfied, it is determined that the power of the nonlinear distortion is not reduced any more and this routine is terminated. On the other hand, if the above equation (18) is not satisfied, it is determined that there is a possibility that the power of non-linear distortion may be reduced, and the routine proceeds to step 130.

  In step 130, the number of corrections L is incremented, the process proceeds to step 110, and the same processing as described above is repeated.

  As described above, in this embodiment, since the sound source signal is directly corrected without modeling the speaker system which is a nonlinear system strictly with a nonlinear function, the nonlinear distortion of the speaker can be reduced with a small amount of calculation.

  In addition, since the recursive processing for further correcting by inputting the corrected sound source signal is performed, nonlinear distortion can be further reduced.

  In addition, for the frequency band where the nonlinear distortion calculated this time is equal to or greater than the previously calculated nonlinear distortion, the correction frequency band selection process for correcting the sound source signal is performed by setting the nonlinear distortion calculated this time to 0. Can be reduced.

  The speaker non-linear distortion reducing apparatus 10 according to the present embodiment can be applied to, for example, a toy that outputs a predetermined sound. When the sound to be output is determined in advance, for example, at the time of manufacture, the output sound is used as a sound source signal, the nonlinear distortion reduction process shown in FIG. 3 is executed, and the corrected sound source signal is stored in the nonvolatile memory 12D. Just store it and ship it. In this case, the microphone 16 is not necessary, and the corrected sound source signal may be output from the speaker 14 during use.

  When the speaker nonlinear distortion reducing apparatus 10 according to the present embodiment is installed in a music playback device such as a CD player, for example, a sound source correction mode is provided, and the sound of the CD is played back as a sound source signal in this sound source correction mode. Then, the nonlinear distortion reduction process shown in FIG. 3 is executed. The corrected CD audio signal is stored in the nonvolatile memory 12D. When the CD is reproduced in the reproduction mode, the corrected sound source signal stored in the nonvolatile memory 12D may be reproduced. In this case, since the output audio signal is not fixed, the microphone 16 is necessary.

(Example)
Next, examples of the present invention will be described. The present inventors conducted a performance evaluation experiment on the nonlinear distortion reduction method described above. (1) Experimental conditions, (2) Principle verification experiment, (3) Frequency band selection effect verification experiment, (4) Reproducibility verification experiment, (5) Sound source dependency verification experiment, (6) Environment A dependency verification experiment will be described.

(1) Experimental conditions The equipment used in the experiment has a configuration in which a sound level meter, a speaker, and an AD-DA converter are connected to a general personal computer. The personal computer as the controller 12 is LATITUDE D630 from Dell, the sound level meter as the microphone 16 is NA-20 manufactured by Lion, the speaker 14 is a product imported by Yoshina (11 W, manufactured in China), an AD-DA converter 18 Used M-Audio Fast Track Pro (manufactured by Avid Technology, Inc.).

The sound sources used were 4 types of sine waves and 4 types of general music. There are four types of sinusoidal sound sources with frequencies of 250 Hz, 500 Hz, 1000 Hz, and 2000 Hz, and the amplitudes are all 0.4. The length of each sine wave is ^{2 16} points, as taper process, has been subjected to both ends half-cosine windows 2400 points.

  There are four types of general music sources: piano, stringed instrument, orchestra, and female vocal. The sound source for general music was cut out in about 10 seconds using a music database for RWC research. The piano used “Classic Music No. 27”, the stringed instrument used “Classic Music No. 14”, the orchestra used “Classic Music No. 3”, and the female vocals used “Popular Music No. 17”. There are two experimental locations: the anechoic room (Kogakuin University Hachioji Campus, Building 9, Room 9-204) and the general room (Room 9-202). The distance between the speaker and the sound level meter is about 60 cm in both the anechoic room and the general room.

(2) Principle Verification Experiment An experiment was performed in which the non-linear distortion reduction processing for the speaker according to the present invention was performed once, and the operation was performed in accordance with the principle and the distortion was reduced. The experimental place was a general room, and the sound source used was a sine wave with a frequency of 2000 Hz. The results are shown in FIGS. The six graphs shown in FIGS. 4 to 9 show the frequency characteristics of the input and output in the course of each operation.

FIG. 4 shows frequency characteristics of the input signal (sound source signal) (variable name X ₀ ).
FIG. 5 shows frequency characteristics of the output signal of the speaker with respect to the input signal (variable name Y ₀ ).
FIG. 6 shows frequency characteristics of the input signal that has been corrected (first time) (variable name X ₁ ).
FIG. 7 shows the frequency characteristic of the output signal of the speaker with respect to the corrected (first time) input signal (variable name Y ₁ ).
FIG. 8 shows the frequency characteristics of the corrected (second time) input signal (variable name X ₂ ).
FIG. 9 shows the frequency characteristics of the output signal of the speaker with respect to the corrected (second time) input signal (variable name Y ₂ ).

  4, 6 and 8, the vertical axis represents the voltage level [dBV], and the vertical axes of FIGS. 5, 7 and 9 represent the sound pressure level [dB]. Moreover, all the horizontal axes of FIGS. 4 to 9 represent the frequency [kHz].

  FIG. 4 shows the frequency characteristics of the sound source signal, and since no distortion component is included, only the component of the sine wave 2000 Hz is displayed.

  FIG. 5 shows the frequency characteristics of the output signal of the speaker with respect to the input signal. In addition to the 2000 Hz component, many harmonics due to nonlinear distortion are included.

  FIG. 6 shows the frequency characteristic of the corrected (first time) input signal, and the distortion is increased as compared with the spectrum of FIG. This is because a distortion component (-D (X, 0) / H) for canceling the distortion is added to the sound source signal (see equation (5)).

  FIG. 7 shows the frequency characteristic of the output signal of the speaker with respect to the corrected (first time) input signal. Compared with FIG. 5 corresponding to the signal before correction, the second harmonic is about 5 dB and the fourth harmonic is about 6 dB. The fifth overtone is reduced by about 11 dB, and the seventh overtone is reduced by about 5 dB. When viewed from the overall level, the fifth harmonic is reduced by 2.9 dB.

  FIG. 8 shows the frequency characteristics of the corrected (second time) input signal. FIG. 9 shows the frequency characteristics of the output signal of the speaker with respect to the corrected (second time) input signal. Compared with FIG. 7 after the first correction, the second harmonic is about 3 dB, the third harmonic is about 2 dB, and the sixth harmonic. Decreased by about 3 dB and the overall level decreased by 3.9 dB. From the above, it was confirmed that the present invention operates correctly in principle.

(3) Frequency band selection effect verification experiment In order to verify the effect of frequency band selection, (A) when frequency band selection is always performed, (B) when frequency band selection is not performed at all, (C) normal case ( An experiment was conducted on selecting a frequency band only when there is an effect. The sound source used is a 2000 Hz sine wave. FIG. 10 shows the transition of the overall nonlinear distortion reduction under the three conditions (A) to (C).

  FIG. 10 shows the level of nonlinear distortion reduction with respect to the number of corrections. The horizontal axis represents the number of times the level was measured (times), and the vertical axis represents the nonlinear distortion level [dB]. Further, the points on the grid are not subjected to the band selection process, and the points between the frequency grids indicate the values after the band selection process and the determination process are performed.

  As shown in FIG. 10, when (A) frequency band selection is always performed, a portion where the level increases at the time of band selection was seen. (B) When frequency band selection is not performed, the level decreases monotonously. (C) In a normal case, the difference between (A) and (B) gradually opens from L = 2, and finally the distortion level is reduced to near 3 dB than in (A) and (B). From this result, it was found that the frequency band selection processing is effective.

(4) Reproducibility verification experiment The present inventors conducted an experiment to verify whether the same distortion reduction effect can be obtained stably even when the nonlinear distortion reduction processing according to the present invention is executed a plurality of times. The experimental place is a general room, and the sound source used is a 2000 Hz sine wave. In this experiment, the same process was performed five times. The experimental results are shown in FIG. The vertical axis and horizontal axis of the graph shown in FIG. 11 are the same as those in FIG.

  As shown in FIG. 11, it has been found that even when the same processing is executed a plurality of times, the same level of nonlinear distortion reduction effect can be obtained. From this result, the reproducibility of the present invention could be confirmed.

(5) Verification experiment of sound source dependency In the above-described (2) principle verification experiment and (3) reproducibility verification experiment, a 2000 Hz sine wave was used as a sound source. In the sound source dependency verification experiment, an experiment was performed in which the sound source was changed in order to verify whether the non-linear distortion could be reduced by other sound sources as well. The sound sources to be used are (1) 8 sound sources of 4 types of sine waves and 4 types of general music shown in the experimental conditions, and the experimental environment is in a general room.

  FIG. 11 shows the final reduced levels for the eight sound sources. The horizontal axis in FIG. 11 indicates the type of sound source, and the vertical axis indicates the amount of reduction in nonlinear distortion. As shown in FIG. 11, although there are variations depending on the sound source, the nonlinear component is reduced in any sound source, and it has been confirmed that the present invention is effective for various sound sources.

(6) Environment-dependent verification experiment In order to investigate the influence of the presence or absence of background noise and reverberation on the amount of nonlinear distortion reduction, the inventors conducted a comparative experiment of nonlinear distortion reduction in an anechoic chamber and a general room. It was. The sound source is the same as that used in (5) the sound source dependency verification experiment. As shown in FIG. 11, when the final reduction levels in the anechoic room and the general room are compared, all the reduction rates in the general room are larger than those in the anechoic room except for the 500 Hz sine wave. It was. However, since there is almost no difference between the two except for the 2000 Hz sine wave, the difference between the anechoic chamber and the general chamber is within the measurement error range, and it is considered that the influence of background noise and reverberation on the performance of the present invention is small. . Thus, it was found that the present invention can be used in a general room and is practically useful.

  As described above, according to the present invention, it was confirmed that the nonlinear distortion was reduced even when the experimental environment and the sound source used were changed. The nonlinear distortion reduction performance was 1 to 11 dB for sine waves and 5 to 9 dB for music.

10 Speaker Nonlinear Distortion Reduction Device 12 Controller 12F Bus 12 Number 14 Speaker 16 Microphone 18 AD-DA Converter

Claims (6)

Measuring means for measuring the impulse response from the speaker to the microphone,
A non-linear component of the output signal is calculated based on a sound source signal output from the speaker, an output signal from the speaker output from the speaker and input from the speaker, and a measurement result of the impulse response. A calculation means;
Correction means for directly correcting the sound source signal based on the nonlinear component and the impulse response;
A non-linear distortion reduction device for speakers with The calculation means corrects the sound source signal corrected by the correction means, the correction sound source signal is output from the speaker and the correction output signal from the speaker is input from the microphone, and the measurement result of the impulse response A process of calculating a non-linear component of the correction output signal based on the correction, and a process of correcting the correction sound source signal again based on the measurement result of the non-linear component of the correction output signal and the impulse response. The non-linear distortion reducing apparatus for a speaker according to claim 1, wherein: is repeated until a predetermined termination condition is satisfied. The calculation unit calculates the nonlinear component for each frequency band, and the correction unit calculates the nonlinear component calculated last time among the nonlinear components calculated for each frequency band. The non-linear distortion reduction apparatus for a speaker according to claim 2, wherein the sound source signal is corrected by setting the non-linear component calculated this time to zero for the frequency band as described above.   The calculation means calculates a linear component that is an ideal frequency component of the output signal based on a value obtained by Fourier transforming the sound source signal, the output signal, and the impulse response, and the linear component and the output signal are Fourier transformed. Based on the converted value, the nonlinear component of the output signal is calculated,
  The correction means directly corrects the sound source signal by subtracting a value obtained by dividing the nonlinear component by a value obtained by Fourier transforming the impulse response from the sound source signal.
  The non-linear distortion reduction apparatus for a speaker according to any one of claims 1 to 3. Measuring the impulse response from the speaker to the microphone;
A non-linear component of the output signal is calculated based on a sound source signal output from the speaker, an output signal from the speaker output from the speaker and input from the speaker, and a measurement result of the impulse response. Steps,
Directly correcting the sound source signal based on the measurement result of the nonlinear component and the impulse response;
A method for reducing non-linear distortion of a speaker including A non-linear distortion reduction program for a speaker for causing a computer to function as the non-linear distortion reduction apparatus for a speaker according to any one of claims 1 to 4 .

Images