JP3472643B2 - Interpolator - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an interpolation device for interpolating the delay time of a delay device.

The delay device using the interpolation device of the present invention is particularly suitable for application to a sound image localization device.

[0003]

2. Description of the Related Art A sound image localization device for locating a sound image in a three-dimensional space is a device for locating a sound source position P with respect to a listening position O in a three-dimensional space as shown in FIG. The sound image smoothly moves in the three-dimensional space based on the coordinate data indicating the position of the sound source sent to the sound source, and the sound image localization is performed.

FIG. 2 shows a block configuration example of this sound image localization apparatus. To localize a sound image in a three-dimensional space,
This can be realized by performing direction localization and distance localization of the sound source P viewed from the listening position O. In FIG. 2, a sound source position input means 1 is used by an operator to input a sound source position to be localized by setting a three-dimensional space on a display screen, for example.
Transfer to. The localization parameter calculation means 2 calculates angle data used for direction localization by the direction localization means 4 and delay time (address value) data used for distance localization by the distance localization means 3 within a period T and supplies them to each. is there.

The direction localization in the direction localization means 4 is a filtering process that simulates a phase difference and a volume difference between the left and right ears, a change in frequency characteristics due to an incident angle to the ears, and a crossing between the left and right ears when the reproducing means is a speaker. The process of canceling the talk is included. The parameter is given as two angles, for example, an angle representing a horizontal direction and an elevation angle, which represent a direction in three dimensions.

The distance localization in the distance localization means 3 is shown in FIG.
As shown in, a path (direct sound) L that directly reaches the listening position O from the sound source P, and a path (reflected sound) L _R that reflects from the sound source to the virtual reflection wall (for example, the floor) and reaches the listening position.
It can be realized by the difference in transmission time. There may be a plurality of virtual reflection walls R, and there may be reflection between the walls. The parameter is given, for example, as a sound transmission time calculated from the distance of each transmission path.

When trying to move the sound image localization in a three-dimensional space with time, the resolution by which a person can recognize the change in the direction localization due to the movement of the sound source is coarse, expressed as an angle, in steps of several degrees. Therefore, a process of holding direction localization data in increments of several degrees in the form of a table, and the direction localization means 4 switches the direction localization data by a method such as a simple crossfade every time the angle data is sent. There is no problem even if you do.

On the other hand, with respect to distance localization, it can be seen that subtle changes can be perceived sensitively from the fact that the transmission time changes as the sound source moves and the Doppler effect can be recognized as a real phenomenon.

The sound source position may be manually moved in real time by using a device such as a mouse or a tablet that outputs coordinates as the sound source position input means 1. Alternatively, the locus of movement may be a function with time as a parameter. Even if it is given by the calculation formula x _p = f _x (t) y _p = f _y (t) z _p = f _z (t), or a table is used to obtain the three-dimensional coordinates that form the shape May be given in order.

However, since the angle of direction localization and the delay time of each path of distance localization must be calculated from the coordinate data, when the calculation is processed by the CPU, the coordinate data is cycled longer than the calculation time. Must be given. On the other hand, when using a table, you may have the result itself calculated by calculation instead of the coordinate data as table data. At that time, the calculation is not necessary and the time required to read the table data is short, so the data read cycle Can be shortened, but the shorter the read cycle, the more table data is required.

For example, the listening position O as shown in FIG.
Consider three-dimensional coordinates as points. Coordinate data of sound source position P
Is given in Cartesian coordinates, the angle of orientation and
To obtain the distance localization delay time, perform coordinate transformation calculation.
You have to. That is, the coordinates of the sound source position P are directly
In cross coordinates (x_p, Y_p, Z_p), In polar coordinates (γ_p,
θ _p, Φ_p), The coordinate change from Cartesian coordinates to polar coordinates.
The following formula γ_p= (X_p ²+ Y_p ²+ Z_p ²)^1/2 θ_p= Tan^-1(Y_p/ X_p) φ_p= Tan^-1[Z_p/ (X_p ²+ Y_p ²)^1/2] Done in.

Further, in the same coordinate system as in FIG. 4, when there is a reflecting wall R that regularly specularly reflects on a plane at a position as shown in FIG. 5, the distance L _R = [(z _p -2z _R ) ² + x _p ² + y _p ² ] ^1/2 angle φ _R = tan ^-1 [(2z _R -z _p ) / (x _p ² + y _p ² ) ^1/2 ] angle θ _R = tan ^-1 (y _p / X _p ) = θ _p , the propagation time and angle of that path must be determined. However, in the above equation, the Cartesian coordinates of the listening position O are (0, 0,
0), the Cartesian coordinates of the sound source position P are (x _p , y _p , z _p ),
The reflection wall R is z = z _R , and x and y are arbitrary.

The above calculation must be executed every time the sound source coordinate data is given.

[0014]

[Problems to be Solved by the Invention]
The delay time given to the input signal is as shown in FIG.
Made with a delay line composed of delay RAM
It In other words, connect the RAM addresses in a ring
Every time, write the input data to this RAM sequentially in time series
And delay the timing of reading that data.
Delay is realized by and. Therefore, the delay time is actually RA
It is given by the value of the read address of M. So, for example,
Distance D [m], sound velocity 340 [m / sec], sampling
F frequency _S[Hz] and write address is fixed
And change the read address, Read address = D × f_S/ 340 + write address Must also be calculated.

Here, the address value is discretely given as a calculation result every time the coordinate data of the sound source is given, and its period is set to T.

When the delay time is changed by such a RAM, as shown in FIG. 8 (a), the address (that is, the delay time) changes if the RAM is directly read with the address value of the RAM which is discretely given at each time period T. Waveform becomes discontinuous, and the pitch change caused by the Doppler effect is given as a derivative of this waveform, so that it is heard as pulse noise at each periodic point as shown in FIG. 8B. For this reason, in order to change the delay time smoothly, it is necessary to read out with a finer resolution than the actual address. The method includes linear interpolation, for example.

FIG. 9 (a) shows this linear interpolation. As shown in the figure, the address value interpolated at each interpolation point in the address input cycle T is generated so that the sequentially input address values are connected by a straight line. However, even in this case, as shown in FIG. 9B, the differential value (that is, the pitch) of the change waveform of the address value changes in a stepwise manner, and the Doppler effect sound due to the movement of the sound source is inaudible.

Therefore, based on the address values given discretely as shown in FIG. 6A, interpolation is performed between them, and a smooth curve is formed on the time axis as shown in FIG. 6B. It is desirable to create a changing address value, and a calculation such as a spline function can be considered for this, but in this case, the calculation is complicated and data accuracy during the calculation is considerably required. Further, although interpolation filter data may be stored in a table for processing, when the cycle T becomes large, the number of taps becomes large and the amount of memory required becomes very large.

The present invention has been made in view of the above problems, and provides an interpolating device that interpolates between discretely input data with a natural feeling to create output data with a simple circuit configuration. The purpose is to

[0020]

FIG. 1 is a diagram for explaining the outline of the present invention. In order to solve the above problems,
The interpolating device according to the present invention is, in an interpolating device for controlling the delay time of a delay device, data input means 101 for discretely and sequentially inputting data representing the delay time, and new data input by the data input means 101. Then, the difference calculation means 102 for calculating the difference between the input data and the output data, and the addition data for performing interpolation by performing filtering processing and coefficient multiplication processing on the difference calculated by the difference calculation means 102 Addition data operation means 103 for sequentially calculating at each interpolation point in the data input time interval, and accumulation means 1 for adding the addition data from the addition data operation means 103 to the output data to sequentially create new output data.
04, and the filtering process of the addition data calculation means 103 is such that the change width of the addition data is changed smoothly so that the time differential value of the output data changes smoothly.

As another form of the interpolating apparatus according to the present invention, in the above-mentioned interpolating apparatus, a timer means for monitoring whether the latest data input interval by the data input means exceeds a predetermined time, and an input by the data input means. Second difference calculation means for calculating a difference between the generated data and the output data;
And a limiter means for inputting an output value of the difference calculating means and limiting the input value to a predetermined value when the input value exceeds a predetermined value, and outputting the added data as the addition data. When it is detected that the difference is calculated, the difference calculated by the second difference calculation means is input to the limiter means, and the accumulation means outputs the addition data from the limiter means instead of the addition data from the addition data calculation means. Configure to add to the data.

[0022]

In the former form of the interpolating device, the addition data created by the addition data calculating means 103 is such that the change width is smoothly changed so that the time differential value of the output data changes smoothly, and the output data is naturally changed. It can be interpolated with a feeling.

Further, according to the interpolating device of the latter form, when the data input is interrupted, for example, it is detected by the timer means, and the last input data and the output data are output by the second difference calculating means. Since the difference of is calculated and given to the accumulating means through the limiter means as the addition data, even if the convergence of the output data deviates from the target value (the last input data), this is gradually added to the end. You can get close to the input data.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. In this embodiment, a case will be described in which the interpolation device of the present invention is applied to the control of the delay line in the distance localization means in the sound image localization device. FIG. 3 shows an example of the configuration of such distance localization means. In FIG. 3, an input signal is input to each of a delay line 31 that simulates a direct sound path and a delay line 32 that simulates a reflected sound path. The delay lines 31 and 32 are composed of delay RAMs in which addresses are connected in a ring shape. The address interpolators 33 and 34 give read addresses to the delay lines 31 and 32, respectively.
At each interpolation point obtained by dividing the input time interval of address values given discretely in time into a plurality of interpolation points, an interpolation address for smoothly connecting the input and output address values is generated and output as a read address.

FIG. 7 is a diagram showing the concept of the configuration and operation of the address interpolators 33 and 34. In FIG.
301 and 304 are adders for adding two input signals, and 3
02 is a low-pass filter (L
PF), 303 is a multiplier for multiplying the input signal by a coefficient K,
305 and 307 are 1-sample delay units, 306 is a sign inverter, and 308 is a limiter that limits an input value to a predetermined value and outputs it. Here, the portion indicated by A in the figure is an accumulating portion for sequentially adding the added value from the multiplier 303 to the output address value held in the one-sample delay portion 305 for each sample, and is indicated by B. The part indicated by 1 is a part that allows the 1-sample delay section 307 to cycle and hold a fixed value.

It is assumed that the angle and the delay address value are calculated each time the sound source position input means gives the sound source coordinate data, and the discrete address value is sent to the address interpolator at every cycle T. The discrete address value is input as operation data, and the value is held as the address value WA until the next new address value is input. Further, this address interpolator is equipped with a detector for detecting whether or not there is a change in the input address value, and a timer for measuring the time, and the contents of calculation are switched depending on the value.

If this address value WA is used as it is as a read address value, the address value changes as shown in FIG. 8A, and the change in pitch is a differentiation of the address value. Since it changes like a pulse as shown in 8 (b) and is heard as noise, it is useless. Here, the value of the vertical axis is arbitrary.

Next, in 1) and 2) of FIG.
Consider the case of linear interpolation without 02. in this case,
The adder 301 obtains a difference WB between the input address value WA and the output address value WC one sampling period before,
This is multiplied by the multiplication coefficient K in the multiplier 303, and is set to 1 in the accumulation section A.
The output data WC before the sampling cycle is accumulated.
Multiplication coefficient K as will be the address value given after a time T _1, is intended to make the amount of change in order to make changes in a straight line towards its value. Therefore, in this calculation, it works so that the space between the adjacent discrete address values changes linearly.

When a new address value is input, the operation 1) is executed. The difference WB between the newly input address value WA and the output address value RC one sample before is taken. The number of samples corresponding to the data WB at time T ₁ : T
_When divided by ₁ × f _S , the value should be changed per sample, so K = 1 / T ₁ × f _S. K is always the same value if the time T ₁ is fixed. By adding this value to the address value RC of the 1-sample delay unit 305, a new current address value WC is generated, and the contents of the 1-sample delay unit 305 are rewritten.

After the next sample, the calculation of 2) is performed until the next input value is given. In this calculation of 2),
Since the difference data WB is held in the holding unit B, the constant addition value K · WB is always added to the address value RC of the one-sample delay unit 305 in the accumulation unit A. The held address value WC is created. As a result, the input address value WA is reached after the time T ₁ and the same operation is repeated when the next new address value is input. In this case, linear interpolation is performed, the address value changes as shown in FIG. 9A, and the pitch change is shown in FIG.
It becomes like (b). Since the pitch changes stepwise, it sounds unnatural as a simulation of sound source movement.

Therefore, the low pass filter 302 is used in the present invention. That is, since the change amount of the address value is a value proportional to the pitch change, the difference WB is set so that the time derivative (that is, the pitch) of the output address value changes smoothly.
When the low-pass filter 302 having a gentle curve is inserted, the pitch change does not occur. In order for the sound source movement to be heard in a natural way, it is only necessary to eliminate steps in the pitch change.
For this, a simple low-pass filter of the first order is sufficient.
In this case, the change of the address value is as shown in FIG. 10 (a), and the change of the pitch is as shown in FIG. 10 (b). Low pass filter 30
If 2 is inserted, the time required to reach the given address value becomes long, so it is advisable to set the multiplication coefficient K and the constant of the low-pass filter 302 after taking that into consideration.

Next, when the new address value is the same as the previous value or is not input, the calculation of 2) continues to be performed, but the difference value WB is calculated only by the calculation of 1) and 2). Since the added value K · WB according to is continuously added to the data RC, the output address value WC continues to change, and the deviation from the target address value WA becomes large. Therefore, in order to stop the change, the calculation of 3) is performed.

In order to perform the operation of 3), when the input data changes, the operation of 1) is executed, so that the value of the timer is reset to 0 there. The operation of 2) is executed from the next sample, but when the time of the timer reaches or exceeds the predetermined time T _t , the operation of 3) is executed.
The value of this predetermined time T _t is set to be equal to or larger than the maximum value of T. The calculation of 3) works so that the difference between the input address value and the output address value approaches 0.

That is, in the operation 3), the difference between the address value data WA input last and the address value data RC one sample before is calculated, and this difference is given by the limiter 308.
And input to the accumulating section A. When the absolute value of the input difference exceeds a predetermined value, the limiter 308 limits this by the predetermined value and outputs it. For example, assuming that the absolute value of the input difference does not exceed the predetermined value d, the limiter 308 outputs d when the input difference X (> d).
When the input difference is -X (<d), it is limited to -d.
Therefore, the change width of the output data in the accumulator A can be suppressed to a small value by the predetermined value. When the current output address value approaches the last input address value WA, the difference output value of the adder 301 becomes smaller and eventually approaches 0. Then, the output data of the accumulator A coincides with the address value WA and does not change. In the calculation of 3), the output address value completely converges to a value equal to the input address value.

In the calculation of 2), the target address value may not be reached or may be exceeded due to the calculation error of the low-pass filter or the coefficient K and the setting of T ₁ and T _t . And at that time, the new address value is the same as the previous value,
It is desirable to correct this error when no input is made (that is, when the calculation in 3) is performed. However, performing the calculation in 3) means that the movement of the sound source has stopped, so if the pitch changes unnaturally at this time, the sound image that should have been localized will move. It is unnatural and not good. Therefore, in the calculation of 3), a limiter is inserted in order to limit the amount of change to a value at which no change in pitch is felt. It is better to set T _t = T ₁ so that the output address value does not change as much as possible in the calculation of 3). Further, if T _t > T, 3) is not executed.

The value of T may fluctuate due to the sound source position input means, and the processing for each cycle T may be slightly delayed due to other calculations. Therefore, the cycle T is not always a constant value. Further, when the time T ₁ is set longer than the cycle, the coefficient K is made smaller, and the tracking of the change of the input address value becomes slower as shown in FIG. If the coefficient K is set too large in order to set T ₁ to be short, the output address value may wavy even if the change in the input address value is constant as shown in FIG. This is very unnatural as a simulation of sound source movement, because the pitch change becomes high or low only at the place where it originally becomes high. Therefore, when there is fluctuation in T, it is preferable that the value of T _{1 be} equal to or larger than the maximum value of T.

In the interpolating apparatus of the present invention, it is necessary to consider so as not to drop data bits as much as possible during calculation. When the change amount of the address value is minute, the difference value WB becomes considerably small. In addition, the low-pass filter 302
Since the bits are lost even in the operation of, the input signal level should be as high as possible. By the way, the multiplier 303
Since the coefficient K to be multiplied in is smaller than 1, the output signal is smaller than the original signal. The order of the low-pass filter 302 and the multiplier 303 may be opposite to that shown in the drawing, but in consideration of the bit loss, it is better to place the low-pass filter 302 before the multiplier 303 that multiplies the coefficient K.

As described above, the required address accuracy can be obtained by a simple calculation, and the address changes smoothly even if the period T is increased, so that the amount of coordinate data can be reduced.

In a transmission path that is not so important, it is possible to simply interpolate the input address using only the low pass filter instead of the address interpolating device of the present invention. Changes in address value and pitch are as shown in FIGS. FIG. 12 shows the case where the time constant of the low-pass filter is longer than that of FIG. In this way, if the time constant of the low-pass filter is lengthened, the unnaturalness of the pitch change is reduced, but the tracking of the change of the input address value is delayed.

In the low pass filter for simplification described above, if an attempt is made to create a low pass filter with a long time constant,
Although coefficient precision and data precision are considerably required, the time constant becomes n times when the 1-sample delay portion of the 1st-order IIR type low-pass filter as shown in FIG. FIG. 15A shows a configuration example in which n = 4.

However, since unnecessary harmonics are generated in this configuration, if an FIR type filter for adding n taps to 1 / n and adding them as shown in FIG. Can be reduced. Since the address value which is the input signal is a step wave having a period T, there are originally few harmonics due to the aperture effect, and attenuation of this level is sufficient. As shown in FIG. 18, the waveform of this low-pass filter seen on the time axis changes stepwise at n-sample periods by the n-sample low-pass filter in the portion where the input address value is held at a constant value. Waveform and n tap FI
The R filter is used to connect the steps of the staircase with straight lines.

Therefore, it is possible to realize a low-pass filter having a long time constant for interpolating a staircase wave with a simple algorithm such as the low-pass filter shown in FIG. 17 even if the accuracy of data and coefficients is low. The low-pass filter of FIG. 17 has a configuration in which an n-sample primary filter and an n-tap FIR filter are combined.

In the localization parameter calculation means, a function such as a square root appears as shown in the above example. Since such a function requires more calculation time than the four arithmetic operations, if the number of such calculations increases, it becomes impossible to shorten the cycle T. Therefore, we will describe a method of shortening the calculation time when calculation is performed by the recurrence formula of the function.

For example, the recurrence formula of the square root is S = (X / S + S) / 2, the initial value is appropriately given, this calculation is repeated, and S when the value of S does not change is X. Becomes the square root of. Since the calculation accuracy is not so required during the movement of the sound source position, the calculation of the parameters is finished within a predetermined number of times even if the calculation result of the recurrence formula is not converged so as to be completed within the required time.

When the movement of the sound source is stopped, it is not necessary to newly calculate the parameters, so the calculation is repeated until the result of the recurrence formula converges so that the calculation accuracy can be obtained.
Further, when the moving speed of the sound source is not so fast, the calculation result does not change so much, and the number of calculations decreases when the previous calculation result is set to the initial value so that the recurrence formula converges quickly.

Next, the input switching process of the sound image localization device will be described. In the sound image localization apparatus, if the position of the sound source is suddenly changed and the distance between the listening position and the sound source changes, an abnormal sound is generated as a progress sound of the change in the sound source localization. For example, there is a case where one sound source localization device is used and a plurality of input signals which are not continuous with each other are switched and used, which cannot be used.

For example, it is assumed that the signal is first reproduced and then the signal is reproduced, and the switching is performed almost instantaneously. Also, the signal moves from coordinate a to coordinate b,
It is assumed that the signal moves from the coordinate c to the coordinate d. Since the coordinates b and the coordinates c are not necessarily the same, the coordinates also change instantaneously when the signal is switched from the signal to the signal, and an abnormal sound may occur, and such usage cannot be performed. . Here, an apparatus having a muffling means for solving such a problem will be described.

As shown in FIG. 19, the transmission time is actually given by the value of the delay RAM address, and the difference between the write address and the read address is changed. Before the processing, the multiplication coefficient K is set to a value other than 0.

When switching the input signal, if the muffling process is performed in the following order, an abnormal sound due to a large pitch change does not occur. Process A 1) The multiplication coefficient K is set to 0. 2) Change the read address value. 3) Switch the input signal. 4) Return the multiplication coefficient K to its original value. The order of 2) and 3) may be reversed.

When the multiplication coefficient K is set to 0 or returned to the original value, if it is changed instantaneously, a harmonic wave with respect to the input signal is generated and an abnormal sound is generated. Therefore, the harmonic wave is gradually reduced so as to decrease. It is better to change it. The same applies to switching of input signals.

Here, it is assumed that the signal having the delay time T _d1 is reproduced until the time T _e1 and the signal having the delay time T _d2 is reproduced from the time T _s2 . At this time, at what time, which signal
FIG. 20 shows which delay time is output. Assuming that the process A is not performed, the time T
_An abnormal sound is generated in the output between _e1 and T _s2 .

In this case, as shown in FIG. 20, the sound of the signal is produced until the signal delayed by the delay time T _d2 is output after the mute by the address value switching. If this causes a problem, as shown in FIG. 21, the processing may be performed in the following order.

Process B 1) The multiplication coefficient K is set to 0. 2) Change the read address value. 3) Switch the input signal 4) Wait for delay time T _d2 . 5) Return the multiplication coefficient K to the original value. The order of 2), 3) and 4) may be reversed. As a result, as shown in FIG. 21, the sound is muted at time T _e1 and a signal sound is output after a time (T _d2 + T _s2 −T _e1 ).

When it is not desired to generate the pitch change due to the Doppler effect, the address value is switched by crossfade. In this case, however, since the abnormal sound due to the large pitch change is not generated, there is no need to mute when switching the address value. However, since the signal of the delay time T _d2 is output as shown in FIG. 20 after the signal is switched as it is, if it causes a problem, the sound must be silenced by the method of the process B.

If it is not desired to wait for the delay time T _d2, the following process may be performed. Process C 1) The multiplication coefficient K is set to 0. 2) Change the read address value. 3) Switch the input signal. 4) Erase all data on the delay line. 5) Return the multiplication coefficient K to the original value. Note that 2) may be brought after 4). This process C
Then, even if there are many outputs with different delay times that must be muted, all outputs can be processed in common.

If it is assumed that the switching process is completed within the signal delay time and no noise is generated when the data on the delay line is erased, the following process is sufficient. Process D 1) Switch the input signal 2) Erase all the data on the delay line. 3) Change the read address value.

In the above description, the case where the interpolating device of the present invention is used for interpolating the delay time modulation signal of the delay means (delay line) in the sound image localization device has been described, but the present invention is not limited to this. Not, for example,
A delay device such as an effector (effect device) or a reverberation device including delay means can be similarly used for interpolation of a signal for delay time modulation. Further, even an effect device other than the delay device can be used for the interpolation of the modulation signal for a device requiring a modulation signal which changes smoothly. For example, it can be used for musical tone modulation of electronic musical instruments (vibrato, tremolo, etc.).

[0058]

As described above, according to the present invention, it is possible to change the delay time of the delay device with a natural audible feeling, and by using the limiter means or the like, this output data can be used as a target. Even when it deviates from the value, it can be gradually converged to the value for the purpose.

Further, if it is used as an address interpolating device for performing delay control for the delay line of the distance localization device in the sound image localization device, it is possible to simulate the pitch change caused by the Doppler effect when the sound source is moved with a natural feeling. .

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a diagram illustrating a configuration concept of a sound image localization device.

FIG. 3 is a diagram showing a configuration example of distance localization means in a sound image localization device.

FIG. 4 is a diagram illustrating coordinates of a sound source in a three-dimensional space.

FIG. 5 is a diagram illustrating a reflection path from a sound source to a listening position in distance localization.

FIG. 6 is a diagram illustrating a time change of a read address value given to a delay line.

FIG. 7 is a diagram illustrating the concept of the configuration and operation of the device when the interpolation device of the present invention is applied to the address interpolation device of the distance localization device.

FIG. 8 is a diagram illustrating a case where a discrete address value is directly applied to a delay line.

FIG. 9 is a diagram illustrating a case where an address value is given to a delay line by linear interpolation.

FIG. 10 is a diagram illustrating a case where an address value is given to a delay line by the apparatus of the embodiment.

FIG. 11 is a diagram illustrating a case where an address value is given to a delay line by a simple method using only a low pass filter.

12 is a diagram showing an example in which the time constant is lengthened in the case of FIG.

FIG. 13 is a diagram illustrating a convergence state of interpolation data due to a difference in multiplication coefficient K.

FIG. 14 is a diagram illustrating an example of a first-order low-pass filter.

FIG. 15 is a diagram illustrating an example of a 4-sample first-order low-pass filter.

FIG. 16 is a diagram illustrating an example of a 4-tap FIR filter.

FIG. 17 is a diagram illustrating an example of a low frequency filter.

FIG. 18 is a diagram illustrating a waveform on a time axis of a low frequency filter.

FIG. 19 is a diagram illustrating a simulation of transmission time.

FIG. 20 is a diagram illustrating a state of an output signal when a signal is switched by the process A.

FIG. 21 is a diagram illustrating a state of an output signal when signals are switched by processes B, C, and D.

[Explanation of symbols]

1 Sound source position input means 2 Localization parameter calculation means 3 direction localization means 4 Distance localization means 31, 32, delay line 33, 34 Address Interpolator 301, 304 adder 302 Low pass filter 303 coefficient multiplier 305,307 1 sample delay section 306 sign inverter 308 Limiter

Claims (2)

(57) [Claims] 1. An interpolator for controlling the delay time of a delay device, wherein data input means for discretely and sequentially inputting data representing the delay time, and new data input by the data input means, Difference calculation means for calculating the difference between the input data and the output data, and addition data for performing interpolation by performing filtering processing and coefficient multiplication processing on the difference calculated by the difference calculation means for each interpolation in the data input time interval. The addition data calculation means for sequentially calculating points, and the accumulation means for adding the addition data from the addition data calculation means to the output data to sequentially create new output data, the addition data calculation filtering An interpolation device, wherein the processing smoothly changes the change width of the addition data so that the time differential value of the output data changes smoothly. 2. A timer means for monitoring whether or not a latest data input time interval by the data input means exceeds a predetermined time, and a second means for calculating a difference between data input by the data input means and output data. The timer means further comprises difference calculating means and limiter means for inputting an output value of the second difference calculating means and limiting the input value to a predetermined value when the input value exceeds a predetermined value and outputting the added data. When it is detected that the data input interval exceeds the predetermined time, the difference calculated by the second difference calculating means is input to the limiter means, and the accumulating means adds the added data from the added data calculating means. The interpolation device according to claim 1, wherein the addition data from the limiter means is added to the output data instead of the above.