JP4963356B2 - How to design a filter - Google Patents

Description

  The present invention relates to an approximate filter creation method, and more particularly to a filter design method that has substantially the same characteristics as a base filter and has a smaller number of taps than the base filter.

  2. Description of the Related Art Conventionally, in a sound reproduction system, there has been known a sound field expansion technique that allows a sound image localization direction to be wider than one position of left and right speakers by applying signal processing to sound signals supplied to left and right speakers. Yes. For example, in the technique disclosed in Non-Patent Document 1, by convolving a transfer function (HRTF) from the sound source to the positions of the left and right ears of the listener into the acoustic signals of the left and right channels, the sound source position is as if the listener. The sound field can be perceived as if the sound is coming from.

Non-Patent Document 2 discloses a technique for canceling crosstalk between left and right channels when sound is reproduced by a speaker. By combining such a crosstalk cancellation technique with the technique disclosed in Non-Patent Document 1, the sound field expansion technique of Non-Patent Document 1 is used more effectively while canceling the crosstalk of the left and right speakers. The localization direction can be expanded.
DBAnderson et al, "The sound dimension", IEEE SPECTRUM, March 1997. MRShcroeder et al, "Comparative study of European concert halls: correlation of subjective preference with geometric and acoustic parameters", J. Accoust. Soc. Am., Vol. 55, No. 4, October 1974.

  The filter coefficient used for the signal processing of such a sound reproduction system is, for example, an impulse for each frequency band by performing delay processing and amplification processing on the band pulse obtained by passing the impulse through the band pass filter for each frequency band. The response is calculated, and the impulse response of each frequency band is added so as to cover the entire frequency band.

  The impulse response obtained as described above has a very large number of taps because the number of taps increases during delay processing and addition processing. When such a filter with a large number of taps is used, the number of calculations increases in proportion to the number of taps of the filter, which causes an excessive burden on the CPU of the signal processing device.

The present invention has been made in view of the above problems, and its object is to have substantially the same frequency characteristics as the reference filter and to reduce the burden on the signal processing apparatus, compared to the reference filter. to provide a way to design a small approximation filter with taps.

According to the filter designing method of the present invention, the input signals of the left and right channels are passed through a filter having band pass characteristics set in advance for each of a plurality of predetermined frequency bands, and the filter is passed from the input signals of the left and right channels. A method of designing the filter of an acoustic signal processing apparatus that subtracts a signal of another channel that has passed and outputs a signal resulting from the subtraction as a signal of a corresponding channel of the speaker, and includes a plurality of frequency bands i (i = 1, ···, N; N is the steps of the impulse response [delta] _i of the band-pass filter BP _i corresponding to s each band number) and the first test signal Sm _i, of the first test signal Sm _i a step of the second test signal Sc _i inverts the phase, and inputs the first test signal Sm _i on one channel of said speaker, said 2 test signal Sc _i, inputted to the other channel of the speaker through a time delay regulator and level regulator acquires the measurement signal SL _i, SR _i and picking up sound the speaker is generated in stereo microphone step a, the first test signal Sm _i or the second test signal Sc _i reference signal a sound source for generating sound corresponding signal representing the sound when installed on the left or right position than the speaker SL to ^* _I and SR ^* _i , and the measurement signal SL _i and SR _i approximate to the reference signals SL ^* _i and SR ^* _i , respectively. and adjusting the level, the steps of the adjusted time delay and adjusting the time delay tau _i, the first test signal Sm _i of levels the adjustment A step of the gain to adjust the gain k _i, the steps of the impulse response hc _i delayed by the adjusted time delay tau _i with applying the adjustment gain k _i to the impulse response [delta] _i, the impulse response hc _i The step of adding the impulse response hc ′ for all frequency bands, and the impulse response hc ′ as an unknown system, the unknown system is defined by an adaptive filter whose initial value is a filter having a smaller number of taps than the impulse response hc ′. A step of identifying, and a step of using the identified adaptive filter as the filter hc.

According to the present invention, it is possible to design an approximate filter that has substantially the same characteristics as the reference filter and has a smaller number of taps than the reference filter. By using such an approximate filter as a filter of the sound reproduction system, the processing load of the signal processing device can be reduced.

  Hereinafter, an embodiment of a filter designing method of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram illustrating an acoustic reproduction system that is a target of filter creation. As shown in the figure, the sound reproduction system 10 includes an acoustic signal processing device 30 and left and right speakers 20L and 20R connected to the acoustic signal processing device 30. The acoustic signal processing device 30 includes an input terminal 32 (32L, 32R), a delay circuit 34 (34L, 34R), an arithmetic output device 36 (36L, 36R), and a sound field adjustment filter to which acoustic signals of left and right channels are input 38 (38L, 38R).

  The input signal input to the left channel input terminal 32L of the acoustic signal processing device 30 is supplied to the right channel calculation output unit 36R through the sound field adjustment filter 38L, and the calculation output unit 36R passes through the delay circuit 34R. A signal obtained by subtracting (or adding by inverting the phase) the signal of the left channel that has passed through the sound field adjustment filter 38L is output from the signal of the channel. Similarly, the input signal input to the right channel input terminal 32R is supplied to the left channel arithmetic output device 36L through the sound field adjustment filter 38R, and the arithmetic output device 36L passes the delay circuit 34L to the left channel signal. Then, the signal of the right channel that has passed through the sound field adjustment filter 38R is subtracted (or added by inverting the phase) and output. Output signals from the arithmetic output devices 36L and 36R are D / A converted and supplied to the speakers 20L and 20R.

  In such an acoustic signal processing device 30, the number of calculations in the sound field adjustment filters 38L and 38R increases in proportion to the number of taps of the filter coefficients set in the sound field adjustment filters 38L and 38R. Burden. For this reason, in order to reduce the burden on the acoustic signal processing device 30, it is desired that the number of taps of the filter coefficients of the sound field adjustment filters 38L and 38R is small.

  The design method of the sound field adjustment filter of the present embodiment calculates an impulse response having optimum acoustic characteristics as the sound field adjustment filter 38 (38L, 38R) of the acoustic signal processing device 30 described above, and the impulse response is calculated as an unknown system. As described above, by identifying with an adaptive filter, an approximate filter having substantially the same characteristics as the impulse response and having a smaller number of taps than the impulse response is characterized.

Hereinafter, a method for designing the sound field adjustment filter 38 will be described.
FIG. 2 is a flowchart showing a procedure for designing the sound field adjustment filter 38. 3-10 is a figure for demonstrating each procedure of FIG. First, in step 100 of FIG. 2, as shown in FIG. 3, for example, a narrow-band bandpass filter BP _i of about ¼ octave (i is a number representing a band, i = 1 to N; N is a band) calculate an impulse response of a path number of filters), to their impulse response to the first test signal Sm _i. The center frequency fc _i , the bandwidth fΔ _i , and the number N of each band pass filter BP _i are set so as to cover a frequency region (for example, 1000 Hz to 3000 Hz) that is a target of acoustic signal processing. In the present embodiment, for example, a linear phase FIR bandpass filter is used as the bandpass filter BP _i .

In next step 102, as shown in FIG. 4, and the phase inverting each first test signal Sm _i in opposite phase, the second test signal Sc _i.
Next, in step 104, as shown in FIG. 5, with the first test signal Sm _i is input to the speaker 20L for the left channel speaker 20, regulator 50 and the level adjuster delays the second test signal Sc _i Time The sound is input to the right channel speaker 20R through 52, and the sound generated from the speaker 20 is picked up by the dummy head microphone 54 installed in front thereof, and the measurement signals are SL _i and SR _i . The dummy head microphone 54 is a microphone that can measure the sound pressure at the positions of the left and right ears of a person.

Next, at step 106, the time difference and level difference between the left and right ear measurement signals SL _i and SR _i by the dummy head microphone 54 are measured when a single speaker is installed at a position on the left side of the left speaker 20L. that the left and right signals (hereinafter, referred to as a reference ^signal) _{SL * i,} ^SR * i to best approximate the time difference and the level difference, time delay of the second test signal Sc _i in delay adjuster 50 and a level adjuster 52 times And adjust the level. Thus the adjusted time delay and adjusting the time delay tau _i, also, the maximum value Mc _i of the second test signal Sc _i adjusted, the ratio between the maximum value Mm _i of the first test signal Sm _{i (Mc i} / Mm _i ) it is the adjustment gain _{k i.} For reference signals SL ^* _i and SR ^* _i , as shown in FIG. 6, for example, a reference speaker 56, which is a normal type speaker with independent left and right channels, is installed on the left side of the dummy head microphone 54. It assumed to be previously measured by inputting the first test signal Sm _i to the speaker 56. The reference signal ^SL _* i to reference speaker 56 inputs the second test signal Sc _i, may be measured ^SR _{* i.}

Note that the first test signal Sm _i, by convoluting impulse responses of binaural is inverse Fourier transform of the transfer function HRTF from the left position just beside the listener for the relevant frequency band up to the listener's head The same signal as when the sound source is placed directly beside the left can be obtained, and this signal may be used as the reference signals SL ^* _i and SR ^* _i .
Further, in the above description, the sound source is assumed to be disposed directly beside the left side. However, the present invention is not limited thereto, and may be disposed at a position on the left side or the right side of the speaker 20 according to the sound image localization direction to be expanded.

Next, at step 108, the adjustment time delay τ _i and the adjustment gain k _i of each frequency band i obtained at steps 104 and 106 described above are equal in time delay and gain in two or more adjacent frequency bands. In the case of the allowable error range (for example, ± 10%), these frequency bands are integrated into one band, and common adjustment time delay and adjustment gain values are used. For example, when the adjustment time delays τ _s , τ _{s + 1} and the adjustment gains k _s , k _{s + 1} coincide with each other in the band s and the band (s + 1), as shown in FIG. A bandpass filter having a characteristic capable of covering the passband of the bandpass filter of each of the bands s and (s + 1). When the bands are integrated in this way, the band number i is reassigned with the integrated band as one band.

Next, in step 110, the impulse response δ _i is calculated through the band-pass filter BP _i of each band i.
Next, at step 112, as shown in FIG. 8, the impulse response δ _i of each band is delayed by the adjustment time delay τ _i of the corresponding band _i and multiplied by the adjustment gain k _i to obtain an impulse response hc _i .
In step 114, as shown in FIG. 9 to determine the one of the impulse response hc' adding together all impulse response hc _i.

  Here, the number of taps of the impulse response hc ′ calculated in this way is generally very large. Therefore, in step 116, in order to reduce the number of taps of the obtained impulse response, an approximate filter having substantially the same characteristics as the impulse response is created by the approximate filter creation system shown in FIG.

As shown in FIG. 10, the approximate filter creation system 60 includes a noise generator 61, a filter device 62, an adaptive filter device 64 having an adaptive filter 66, and a delay processing device 63.
The noise generator 61 is a device that outputs white noise u (k). The white noise u (k) output from the noise generator 61 is input to the filter device 64 and the delay processing device 63.

  The filter device 62 outputs a target signal d (k) (corresponding to the first signal in the claims) by performing a filtering process on the input white noise u (k). The impulse response hc ′ (k) calculated as described above is set as a filter coefficient used for this filter processing.

  The delay processing device 63 performs delay processing on the input white noise u (k). The delay time Δk of the delay processing is set such that the sum of the delay time due to signal processing of the adaptive filter 66 described later and the delay time due to the delay processing device 63 is equal to the delay time due to the filter device 62.

  The adaptive filter device 64 includes an adaptive filter 66 and an arithmetic processing unit 65. The adaptive filter 66 can set a filter coefficient with a desired number of taps as an initial value, and sets a filter coefficient with a tap number that does not impose a burden on the CPU of the sound processing device 30 as an initial value.

  The adaptive filter device 64 receives the white noise u (k + Δk) delayed by the delay processing device 63 and the target signal d (k) output from the filter device 62. The adaptive filter device 64 calculates the reference signal y (k) (corresponding to the second signal) by performing signal processing on the input white noise u (k + Δk) in the adaptive filter 66, and further in the arithmetic processing unit 65. An error signal e (k) that is a difference between the reference signal y (k) and the target signal d (k) is calculated. Then, the adaptive filter device 64 updates the adaptive filter hc (k) so that the error signal e (k) is minimized by an LMS algorithm or the like. By repeating this updating process until the filter coefficient of the adaptive filter device 64 becomes substantially constant, the adaptive filter 66 has substantially the same characteristics as the filter coefficient hc ′ (k) of the filter device 62.

  Further, even if the filter coefficient of the adaptive filter 66 is updated, the number of taps does not change. Therefore, the adaptive filter 66 in which the filter coefficient hc (k) is substantially constant has substantially the same characteristics as the filter device 62 having the filter coefficient hc ′, and the number of taps is the filter coefficient of the adaptive filter 66. The number of taps is equal to the initial value of. The approximate filter creation system 60 outputs the filter coefficient hc (k) of the adaptive filter 66 as an approximate filter.

  Returning to FIG. 2 again, in step 118, the phase delay time T up to the peak value of the obtained approximate filter hc is calculated. This phase delay time T is set as the delay time of the delay circuit 34. Further, the obtained approximate filter hc is set as a filter coefficient of the sound field adjustment filter 38.

As described above, according to the approximate filter creation system 60, it is possible to create an approximate filter having the same characteristics as the impulse response and a small number of taps. Thereby, according to the filter design method of this embodiment, the load on the arithmetic processing unit of the acoustic signal processing device 30 can be reduced by using this approximate filter as the filter coefficient of the sound field adjustment filter 38.
The sound reproduction system 10 which is a filter creation target by the filter creation system 40 of the present invention is a commercially available sound reproduction system having a filter processing function. Especially for left and right speakers, even if the listener is laterally offset from the center line of the left and right speakers, by designing an optimal filter for each channel, It was experimentally confirmed that a similar sound field can be reproduced.
In particular, in the case of a sound reproduction system in which the position of a listener such as a car stereo is extremely close to one speaker, if the filter processing is performed only on the channel of the close speaker, the speaker placement Can reproduce a sound field similar to that of a symmetrical sound reproduction system. In such a case, a filter need only be provided on one channel.

  In the adaptive filter device 64 described above, the algorithm for identifying the filter coefficient is not limited to the LMS algorithm, and other algorithms may be used, or an algorithm having a slow convergence may be used. The approximate filter creation system 60 used in the filter design method of the present invention is used not only for designing a sound reproduction system filter but also for designing a filter such as a television or a mobile phone having a filter for acoustic signal processing. Can do.

Here, the impulse response before the creation of the approximate filter in the filter design method described above (that is, step 114 in FIG. 2) was compared with the approximate filter created by the approximate filter creation system based on this impulse response. So I will explain.
FIG. 11A is a diagram showing an example of the waveform of the approximate filter created by the approximate filter creation system, and FIG. 11B is the waveform of the impulse response that is the basis of the approximation. As shown in the figure, an impulse response having 4000 or more taps is approximated by an approximate filter having 256 taps.

  FIG. 12A is a diagram illustrating a waveform of 100 steps including the peak of the approximate filter, and FIG. 12B is an enlarged view of the waveform of 100 steps including the peak of the impulse response. 12A and 12B, it can be seen that the approximate filter approximates the impulse response waveform with high accuracy.

  FIG. 13A is a graph showing the frequency characteristics of the approximate filter, and FIG. 13B is a graph showing the frequency characteristics of the impulse response. By comparing FIGS. 13A and 13B, it can be confirmed that the approximate filter has substantially the same frequency characteristics as the underlying impulse response.

  Furthermore, when an experiment was conducted in which the basic impulse response and the approximate filter were actually auditioned and compared, the subject could not distinguish between the two. Thereby, according to the approximate filter creation system of the present embodiment, it was confirmed that an approximate filter having the same characteristics as the impulse response and having a small number of taps can be created.

It is a figure which shows the sound reproduction system used as the object of filter preparation. It is a flowchart which shows the procedure which designs a sound field adjustment filter. It is a figure for demonstrating the procedure which acquires a 1st test signal from the impulse response of the band pass filter of each band in the design method of a sound field adjustment filter. It is a figure for demonstrating the procedure which acquires a 2nd test signal from a 1st test signal. It is a figure for demonstrating the measurement procedure of measurement signal SLi, SRi. It is a figure for _{demonstrating} the measurement procedure of reference signal SL ^* _i , SR ^* _i . It is a figure for demonstrating the integration process of a zone | band. It is a diagram for explaining a procedure for obtaining an impulse response hc _i from the impulse response [delta] _i of the band-pass filter for each band i. It is a figure for demonstrating the procedure which calculates | requires the impulse response hc which is the characteristic of a sound field adjustment filter from the impulse response hc _i . It is a figure which shows an approximate filter production system. (A) is a figure which shows an example of the waveform of the approximate filter produced by the approximate filter production system, (B) is a figure which shows the impulse response used as the basis of approximation. (A) is a figure showing 100 steps including the peak of the filter coefficient of an approximate filter, (B) is the figure which expanded and displayed 100 steps including the peak of an impulse response. (A) is a graph which shows the frequency characteristic of an approximate filter, (B) is a graph which shows the frequency characteristic of an approximate filter coefficient.

Explanation of symbols

BP _i band-pass filter Sm _i first test signal Sc _i second test signal SL _i , SR _i measurement signal SL ^* _i , SR ^* _i reference signal τ _i adjustment time delay k _i adjustment gain δ _i , hc _i , hc impulse Response 10 Sound reproduction system 20 (20L, 20R) Speaker 30 Acoustic signal processing device 32 Input terminal 34 Delay circuit 30 Acoustic signal processing device 32 (32L, 32R) Input terminal 34 (34L, 34R) Delay circuit 36 (36L, 36R) Calculation output unit 38 (38L, 38R) Sound field adjustment filter 50 Time delay adjustment unit 52 Level adjustment unit 54 Dummy head microphone 56 Reference speaker 60 Approximate filter creation system 61 Noise generator 62 Filter unit 63 Delay processing unit 64 Adaptive filter unit 65 Calculation Processing unit 66 Adaptive filter

Claims (1)

The left and right channel input signals are passed through a filter having a preset band pass characteristic for each of a plurality of predetermined frequency bands, and the signals of other channels that have passed through the filter are subtracted from the left and right channel input signals. The method of designing the filter of the acoustic signal processing apparatus that outputs the signal resulting from the subtraction as the signal of the corresponding channel of the speaker,
A plurality of frequency bands i (i = 1, ···, N; N is the number of bands) the steps of the impulse response [delta] _i of the band-pass filter BP _i corresponding to s husband the first test signal Sm _i,
A step of the second test signal Sc _i and inverting the phase of said first test signal Sm _i,
Inputs the first test signal Sm _i on one channel of said speaker, said second test signal Sc _i, inputted to the other channel of the speaker through a time delay regulator and level regulator, the speaker Picking up the generated sound with a stereo microphone and obtaining its measurement signals SL _i , SR _i ;
Signals representing sound when a sound source that generates sound corresponding to the first test signal Sm _i or the second test signal Sc _i is installed at a position on the left side or the right side of the speaker is used as a reference signal SL ^* _i , SR. ^* Obtaining as _i ;
Adjusting the time delay and level by the time delay adjuster and the level adjuster, respectively, so that the measurement signals SL _i and SR _i approximate the reference signals SL ^* _i and SR ^* _i ;
Setting the adjusted time delay as an adjustment time delay τ _i ;
A step of the gain for the first test signal Sm _i of levels the adjustment and the adjustment gain k _i,
Multiplying the impulse response δ _i by the adjustment gain k _i and delaying it by the adjustment time delay τ _{i to} obtain an impulse response hc _i ;
The method comprising the impulse response hc' this impulse response hc _i are summed for all frequency bands,
Identifying the unknown system by an adaptive filter having the impulse response hc ′ as an unknown system and a filter having a smaller number of taps than the impulse response hc ′ as an initial value;
And a step of setting the identified adaptive filter as the filter hc.

Images