JPH0937400A - Sound image localization controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To localize a sound image at a prescribed position without any sound image surrounding the prescribed position by applying convolutional operation to a sound source by the use of a coefficient calculated based upon a virtual impulse response obtained by adding a reflected sound preceding a direct sound for a specific time. SOLUTION: Digital and analog audio signals are switched by a switch SW, the selected signal is converted into a parallel signal by a serial-parallel converter 5 and the parallel signal is supplied to right and left channel convolvers 6, 7 and 8, 9. Convolutional operation is applied to the sound source data on a time base by the use of a coefficient corresponding to a prescribed sound image localization position supplied from a controlling sub-CPU 11 to a ROM 10. The processed signals are added by respective adders 12, 13 as right and left channel signals and the added signals are outputted to speakers. The coefficient stored in the ROM 10 is calculated based upon a virtual impulse response obtained by adding a reflected sound of -10 to -13dB preceding the direct sound for 9.1 to 11.3 ms. Since the sound image is localized in the prescribed position by the method, a reproduced sound can be listened as a natural sound.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to an actual transducing method.
The present invention relates to a sound image localization control device that makes a user feel as if a sound image is localized at a desired arbitrary position different from a speaker (speaker / headphone, etc.).

[0002]

2. Description of the Related Art Conventionally, when a stereo signal or a monaural signal is reproduced by using general headphones or a plurality of speakers, in the case of headphones, the sound is muffled in the head or caught around the ears. It's lost or spied
In the case of reproduction, there was a feeling that the sound image seemed to be closer to the target position.

Therefore, in order to avoid such inconvenience,
Various techniques (binaural techniques) for recognizing sounds as if they are heard outside the head or at a desired position by convoluting the impulse responses from the sound source collected in free space to both ears into the reproduced sound source have been proposed. There is.

For example, FIG. 6 shows an example of the prior art.
It is a principle view of the sound image localization control device previously proposed by the applicant of the present application. According to this apparatus, the transfer characteristics cfLx and cfRx at the required localization position x are cfLx and cfRx as the coefficients to be realized by the FIR filter processing.
Is prepared in advance and, for example, the data thereof is prepared in the ROM in advance.

For example, when it is desired to localize the sound source X at a desired position, the transfer characteristic cfLx based on actual measurement prepared in the ROM is transferred to the FIR digital filter, and the signal from the sound source X is convoluted. Then, the signal thus processed is reproduced from the pair of speakers SP1 and SP2 to localize the sound image at a desired arbitrary position.

Further, the configuration of this principle will be described.
Now, the transfer characteristic (frequency response of impulse response) from the speaker SP1 to the left and right ears of the listener M is h1L.
The head-related transfer characteristics h1R to the left ear and the head-related transfer characteristics from the sp2 to the left and right ears are defined as h2L and h2R. Also,
The transmission characteristics to the left and right ears of the listener M when the actual speaker is placed at the intended localization position x are pLx and pR, respectively.
Let x.

The transfer characteristics pLx and pRx are, for example,
It is obtained by arranging a speaker and a human head or a dummy head and a microphone at both ears positions in an anechoic space and subjecting the measured one to appropriate waveform processing. FIG. 7 shows a measurement system for obtaining the transfer characteristics pLx and pRx. A pair of microphones ML and MR are installed on both ears of a dummy head (or human head) DM, and a measurement sound from a speaker SP is output. Received, source sound to recorder RAT (reference data)
refL, refR and the sound to be measured (measurement data) L, R are designed to be recorded in synchronization, and a predetermined waveform processing or the like is performed on the basis of the recorded data. The transfer characteristics pLx and pRx described above are obtained.

Then, these transfer characteristics pLx, pRx
Is guided by the following logic. Now, here, the signals obtained in the left and right ears of the listener M are respectively eL,
If eR, then eL = h1L · cfLx · X + h2L · cfRx · X (1a) eR = h1R · cfLx · X + h2R · cfRx · X (1b)

On the other hand, when the source X is reproduced from the target fixed position, the signals obtained in the left and right ears of the listener M are dL and dR.
Then, dL = pLx · X (2a) and dR = pRx · X (2b).

Now, by reproducing the speakers SP1 and SP2, the signals obtained in the left and right ears of the listener M are sourced from the target position.
If the signal matches the signal when the audio is reproduced, the listener M will recognize the sound image as if the speaker were present at the target position.

That is, the conditions eL = dL, eR = dR
And from equations (1a), (1b), (2a), (2b),
When X is erased, h1L · cfLx + h2L · cfRx = pLx (3a) h1R · cfLx + h2R · cfRx = pRx (3b)

From equations (3a) and (3b), cfL
When x and cfRx are obtained, cfLx = (h2R · pLx−h2L · pRx) / H (4a) cfRx = (− h1R · pLx−h1L · pRx) / H (4b) where H = h1L · h2R− h2L · h1R (4c).

Therefore, if the signal to be localized is processed using the transfer characteristics cfLx and cfRx calculated by the equations (4a) to (4c), the sound image at the target position x can be localized. . Specifically, as described above, cfLx and cfRx corresponding to the number of fixed positions x necessary for the principle configuration of FIG. 6 are created in advance, prepared as a ROM, and the filter coefficient thereof is controlled by the controller. belongs to.

[0014]

However, although the above-mentioned configuration has a considerable improvement effect as compared with the prior art, the impulse response measured in free space is as shown in FIG. There was little reflected sound following the direct sound, the desired effect was not obtained, and the sound image that was still recognized sometimes sounded much closer than the intended position. It is also known that the sound image is increased.

Therefore, taking these points into consideration, it has been proposed to collect the impulse response in a room where good acoustic characteristics can be obtained. However, according to this method, although the above-mentioned problems can be improved, the impulse response sampled as described above is very long, and it is not necessary from a mounting point of view to prepare a coefficient corresponding thereto. Not realistic.

Therefore, in most cases, an impulse response that is scaled with a limited window is prepared, or a long reflective structure is artificially created to enhance its effect. It is considered to be.

However, in the former case, the impulse response is cut off on the way, and the effect is diminished. In the latter case, there is a problem that the sound quality becomes a metallic unnatural sound because the reflective structure is artificially created.

Therefore, the present invention solves the above-mentioned various problems and localizes the sound image further away without the sound clinging to the ear, thereby improving the sound quality more than before. An object of the present invention is to provide a volume localization control device in consideration of downsizing of a circuit.

[0019]

In order to solve the above problems, the volume localization control apparatus of the present invention is a signal processed by a pair of convolvers to which the same sound source is supplied from a pair of spaced transducers. In the sound image localization control device that makes the listener feel that the sound image is localized at an arbitrary position different from the pair of transducers, the signal from the same sound source is set to the set coefficient. A pair of convolvers that perform convolution calculation processing accordingly, and storage means that holds a cancel filter coefficient group calculated as an impulse response based on the head-related transfer function measured at each sound image localization position are designated. The cancel filter coefficient group includes a coefficient supply unit that supplies a coefficient corresponding to a sound image localization position from the storage unit to the pair of convolvers. Prior to the direct sound coming directly to,
It is characterized in that it is a coefficient group calculated based on a virtual impulse response in which a reflected sound that comes in late with the reflected sound is added.

Further, in the sound image localization control apparatus having the above structure, the preceding time of the reflected sound with respect to the direct sound is set to 9.1 msec to 11.3 msec with reference to the maximum level of these sounds, and The maximum level of reflected sound is set to -10 dB to -13 dB with respect to the maximum level of the direct sound.

Further, in the sound image localization control device having the above-mentioned configuration, the reflected sound added prior to the direct sound is converged to zero before the direct sound starts. .

[0022]

Embodiments of the present invention will be described below with reference to the drawings. <Basic Concept> The present invention has been achieved based on an analysis of the factors described below. The reason why the short impulse response pointed out in the above-mentioned problem is heard at a position much closer than the target position is considered to be a large factor that the effect of the reflected sound is reduced by the masking effect.

Among typical human auditory phenomena,
The “hearth effect / preceding sound effect” and the “masking effect” can be mentioned. First, regarding the “hearth effect”, FIG.
This will be described with reference to (a) and (b). FIG. 9A shows the positional relationship of the speakers SP1 and SP2 with respect to the listener M,
FIG. 9B is a diagram showing the angular position ψ where the sound image is localized on the vertical axis and the delay time St of the sound emitted from SP2 with respect to the speaker SP1 on the horizontal axis. Two speaker SPs now
1. Assume that SP2 emits an aperiodic and coherent signal. When the same signal is radiated from both speakers simultaneously and at the same level, a sound image is localized in the front direction of the listener M. If there is a difference in the sound emission time from one of the speakers, the sound image is
Move to Ka side.

In FIG. 9B, the angular position where the speaker SP1 is installed is shown as ψ = 40 °, and the angular position where the speaker SP2 is installed is shown as ψ = -40 °.
As shown in the figure, until the sound emitted from the speaker SP2 is approximately 30 msec with respect to the speaker SP1, the sound image is localized on the side of the speaker SP1 which is emitting the sound first. Further, when it exceeds about 30 msec and reaches a delay of about 50 msec, the sound image is divided into two and can be heard. Thus, it can be seen that the preceding sound has a great effect on hearing. Such a phenomenon is called a hearth effect or a preceding sound effect.

Next, the "masking effect" will be described. The masking effect includes a forward effect and a backward effect. The forward effect is a phenomenon in which a reflected sound is erased due to a direct sound having a large level.

The backword effect is a kind of continuous masking and is a phenomenon in which a preceding sound is suppressed by a sound that arrives later. In other words, when the level of the sound that arrives late is high due to the phenomenon of inaudibility,
Such a phenomenon occurs.

FIG. 10 is a diagram for explaining the phenomenon, and FIG. 9A shows the listener M and the speakers SP1 and SP2.
It is a figure which showed the phenomenon when arrange | positioning as shown in FIG.
The vertical axis indicates the level difference between the speaker SP1 and the speaker SP2 [(level Ls0 of the speaker SP1)-(speaker SP
2 level Lst)] and the horizontal axis represents the delay time St of the sound emitted from the speaker SP2 with respect to the speaker SP1.

The firing sound of the speaker SP2 is the speaker SP.
If the level difference becomes about 15 dB or more with a delay of about 10 ms or more with respect to the firing sound of No. 1, the preceding sound, that is, the firing sound of the speaker SP1 is suppressed (inaudible?). ing. Such a phenomenon is called a backword effect.

In the present invention, when the impulse response has a finite length, the above-mentioned problems are solved by focusing on the above phenomenon. In order to increase the effect of the reflected sound, the reflected sound is directly reflected. This is to suppress the level of the reflected sound that precedes the sound and at the same time precedes it to the extent that it cannot be heard by backword masking by the direct sound.

<Embodiment> Next, an embodiment of the sound image localization control apparatus realized by paying attention to the above-mentioned phenomenon will be described. FIG. 1 is a schematic block diagram of the sound image localization control device.
1 is an input terminal for a digital audio signal, and a switch S
It is connected to the terminal a side of W. Reference numerals 2 and 3 are input terminals for the L channel and the R channel of the analog audio signal, respectively, which are connected to the analog-digital (A / D) converter 4, where the left and right analog audio signals that come in parallel are input. It is designed to convert the audio signal into serial digital data. Reference numeral 5 denotes a serial-parallel converter, which is a circuit for converting serial data selectively received from the switch SW into parallel audio data for left and right.

The serial-parallel converter 5 is connected to convolvers 6 and 7 and convolvers 8 and 9 provided in pairs, and the convolvers 6 to 9 have R
The coefficients corresponding to each sound image localization are supplied from OM10,
A convolution operation is performed. This ROM
The coefficient stored in 10 is obtained by using the measurement system as shown in FIG. 7 based on the logical formula described later in consideration of the masking effect as described above, and is controlled by the control sub CPU 11. A predetermined coefficient is selected,
It is adapted to be supplied to each convolver 6-9.

The adders 12 and 13 have convolvers 6 to 9 respectively.
The signals supplied from the above are generated into signals for the L channel and the R channel, and these signals are output terminals 15, 16
It is configured to output from.

Next, how to obtain each coefficient stored in the ROM 10, which is a feature of the present invention, will be described with reference to FIG. 2 and the above-mentioned FIGS. 6 and 7. FIG. 2 is a diagram showing steps of calculating the coefficient. Measurement of Head Related Transfer Function (hereinafter referred to as HRTF) (Step 101) A pair of microphones ML and MR are installed on both ears of a dummy head (or human head) DM, and measurement is performed from a speaker. Receiving the sound, the source sound (reference data) r on the recorder DAT
efL and refR and non-measurement sounds (measurement data) L and R are recorded in synchronization.

As the source sound XH, impulse sound, white noise, other noises or the like can be used. In particular, from the viewpoint of statistical processing, white noise is
Energy in a continuous tone and over the audio band
Since the distribution is constant, the SN is improved by using white noise.

The position of the above-mentioned speaker is, for example, 0 on the front side.
Predetermined as a degree (°)
It is installed at every 30 degrees (12 points) and continuously recorded for each predetermined time. For example, as shown in FIG. 3A, the direct sound is recorded as a waveform in which the first reflected sound, the second reflected sound, the third reflected sound, ... Are continuous.

Cutout of first reflected sound (step 10
2) The first reflected sound is cut out by applying a window in a predetermined range on the time axis to the waveform shown in FIG. 3A [FIG. 3B].

First reflected sound shaping processing (step 10)
3) An unnecessary band component of the cut out first reflected sound, for example,
BPF (bandpass filter) for dips that occur in high frequencies
To remove. At the same time, the level is set so that the first reflected sound is -10 dB to -13 dB with respect to the level of the direct sound. For example, the maximum value of the sampled data is calculated from the average of two points.

Further, the cut-out first reflected sound is added prior to the direct sound in the next step. At that time, the rear side of the added first reflected sound and the direct sound are added. As shown in Fig. 3 (c), do not overlap the front side of
For example, window processing is performed using a half cosine window [FIG. 3 (d)]. By doing so, it is possible to prevent the reflected sound from interfering with the direct sound and obscuring the sound image.

Addition of the first reflected sound prior to the direct sound (step 104) In step 103, the first reflected sound that has undergone shaping processing including level adjustment and the like is added prior to the direct sound. To do. In this case, the leading time t of the first reflected sound is 9 from the maximum level of the direct sound to the maximum level of the first reflected sound.
It has been proved from the data described later that the level difference at the maximum level is added in the range of 1 msec to 11.3 msec and in the range of 10 to 13 dB, which is also effective from the data described later.

Virtual impulse response (Impulse Respons
e; step of calculating IR (hereinafter referred to as IR) (step 10)
5) In step 101, the synchronously recorded source sounds (reference data) refL, refR, and the virtual measured sound (hereinafter, referred to as the reflected sound added in step 104 prior to the direct sound) , Virtual sound to be measured) L and R are processed on a workstation (not shown).

The frequency response of the source sound is X (S), the frequency response of the virtual measured sound is Y (S), and HR at the measurement position.
When the frequency response of TF is IR (S), there is an input / output relationship shown in Equation 5. Y (S) = IR (S) · X (S) (Equation 5) Therefore, the frequency response IR (S) of the HRTF is: IR (S) = Y (S) / X (S) (Equation 6) Becomes Therefore, the frequency response X (S) of the reference,
The frequency Y (S) of the virtual measured sound is calculated in step 104 above.
The data obtained in step S1 is cut out in a time-synchronized window, expanded into a finite Fourier series by FFT conversion, and calculated as discrete frequencies.
(S) is obtained by a known calculation method.

Virtual IR shaping processing (step 106) Here, the virtual IR obtained in step 105 is shaped.
First, for example, by FFT conversion, the first IR obtained in step 102 is expanded at discrete frequencies over the audio spectrum, and unnecessary bands are removed by BPF. Then, the band-limited IR (S) is inversely transformed, and the IR (in-pass response) is multiplied by a window of, for example, a cosine function on the time axis to perform window processing (becomes the second IR).

Cancellation filters cfLx, cfRx
(Step 107) The cancel filters cfLx and cfRx, which are convolvers (convolutional integration circuits), are calculated by the above equations 4a and 4b.
As shown in, cfLx = (h2R.pLx-h2L.pRx) / H (4a) cfRx = (-h1R.pLx-h1L.pRx) / H (4b) However, H = h1L.h2R-h2L -H1R ... (4c).

Here, the speakers SP1 and SP to be arranged
2 head-related transfer characteristics h1L, h1R, h2L, h2
R and the second IR for each angle obtained by the above step as head-related transfer characteristics pLx, pRx when an actual speaker is placed at the intended localization position x
Substitute (impulse response).

The head-related transfer characteristics h1L and h1R correspond to the positions of the L-channel speakers shown in FIG. 4, and if they are installed at 30 degrees (θ = 330 degrees) from the front to the left, An IR of 330 is used. Head-related transfer characteristic h2L,
h2R corresponds to the position of the R channel speaker shown in the figure, and is, for example, 30 degrees (θ = 30) from the front to the right.
If so, an IR of θ = 30 degrees is used (that is, one that is close to the assumed actual sound image reproduction system is selected).

As the head-related transfer characteristics pLx and pRx, not only the 180 ° range of 90 ° from the front, which is the desired sound image localization, but also in a wide space (entire space) beyond that, for example, By substituting IR for every 30 degrees, cfL of the entire space corresponding to it
The x, cfRx group is determined.

Cancellation filters cfLx, cfR
The x group is finally obtained as IR (impulse response) which is the response on the time axis. The cancellation filters cfLx and cfRx are calculated by the equation 4a as follows. First, a kind of inverse filter H ⁻¹ with respect to H in Expression 4b is obtained by the least squares method, and this is inverse FFT-transformed into a time function h (t). Also, each term h1 of the equation 4a
By expressing L, h1R, h2L, h2R, pRx, and pLx by a time function, respectively, the following equation is established.

CfLx (t) = (h2R · pLx−h2L · pRx) · h (t) (7a) cfRx (t) = (− h1R · pLx + h1L · pRx) · h (t) ... (7b)

Therefore, these equations (7a), (7
The coefficients of the cancellation filters cfLx and cfRx are obtained from b). As is clear from these equations, the coefficients of the cancel filters cfLx and cfRx are made shorter by each head-related transfer characteristic h1L, h1R, h2L, h2R,
It is extremely important to shorten pRx respectively. Therefore, as described above, various processes such as window processing and shaping processing are performed in each step to perform the head-related transfer characteristics h1L and h1.
R, h2L, pRx, pLx, h2R are shortened.

Scaling of the cancellation filter at each localization point x (step 8) In addition, the spectrum distribution of the sound source (source sound) actually processed by the convolver (cancellation filter) is pink when viewed statistically. There are things that are distributed like noise, or things that drop gently in the high range. In any case, the sound source is different from a single sound, so when folding calculation (integration) is performed,
There is a risk of causing a flow and distortion. Therefore,
Cancel filter c to prevent overflow
Of the coefficients of fLx and cfRx, the maximum gain (for example,
Find all of the cancellation filters cfLx and cfRx (the sum of squares of each sample value), and when all the coefficients are convolved with the white noise of 0 dB, all coefficients are scaled. To ring. In practice, the maximum absolute value is 0.1 to the allowable level (amplitude) 1.
It is good to attenuate so as to be in the range of 0.4.

Then, for example, by a cosine window, both ends are set to 0 according to the number of actual convolver coefficients.
Window processing to shorten the effective length of the coefficient.
In this way, the data group that has been scaled and finally supplied to the convolver as coefficients and stored in the ROM 10 (in this example, the coefficient of 12 sets of convolvers capable of sound image localization every 30 degrees). Group) cfLx, cfRx are obtained.

For example, when the sound image localization control device shown in FIG. 1 is used for a game machine, as shown in FIG. A pair of speakers SP1 and SP2 are arranged separately from each other, or a head horn is prepared.

The switch SW is selected depending on whether the signal, which is a sound source such as an airplane sound from the synthesizer, coming through the input terminals 1 or 2 or 3 is an analog or digital signal.
After being selectively switched by the serial-parallel converter 5, the parallel signals are converted into parallel signals, and then the convolvers 6 and 7 are provided in pairs for the left and right channels.
And the convolvers 8 and 9. On the other hand, the control signal from the control sub CPU 11 is stored in the ROM.
10, the coefficient corresponding to each predetermined degree position is extracted based on this signal, and is supplied to each convolver 6-9.

In each of the convolvers 6 to 9, for example, sound source data for airplane sound is subjected to convolution calculation processing on the time axis by a coefficient supplied from the ROM 10, and each adder 1
The signals 2 and 13 are added as left and right signals, and then output from the output terminals 14 and 15. The output signal is then converted into an analog signal by a digital-analog converter (not shown), etc.
Or be supplied to headphones.

In this way, the sound reproduced from the speaker or the headphones is localized so that the crosstalk to both ears of the gamer is canceled and the sound source is located at a desired position. In particular, in the present device, when the speaker is used, the sound source is not localized at a position closer than the intended position unlike the conventional device, and the sound image is located at a desired position. Can be localized. In addition, when headphones are used, the sound is not caught around the ears, and the sound source is localized so that it is located at a distant position. In addition, there is no rise in the sound image, and it can be heard as an extremely natural sound, and it is possible to provide a virtual reality filled with a sense of reality.

In the step 104, the leading time t of the first reflected sound with respect to the direct sound is set to 9.1 mse.
The range of c to 11.3 msec, and the level difference between them at the maximum level is 10 to 13 dB,
It is based on a predetermined experimental result. In this experiment, we obtained the coefficients for dozens of samples by the steps described above,
Various data were obtained by using an apparatus having the same configuration as that in FIG. Further, in this case, the first reflected sound added as the preceding sound is symmetrical to each of the left and right channels. FIG.
Is a diagram summarizing the data of the experimental results, showing the leading time (ms) of the reflected sound in the vertical direction and the level (dB) in the horizontal direction.
Is shown. As is clear from this figure, it can be seen that a remarkable effect is recognized particularly in the above range.

Further, although not shown, according to the verification by the present inventor, it is more effective to make the preceding sound asymmetric rather than symmetric to the left and right channels as in the above experiment. Has been confirmed. Further, in the present embodiment, one reflected sound is added as the preceding sound, but not only one reflected sound but also the second reflected sound, for example, depending on how the reflected sound is generated, reflection time, level, etc. It goes without saying that adding a plurality of sounds such as the third reflected sound and adding a reflected sound other than the first reflected sound can be appropriately applied.

[0058]

According to the present invention, the sound reproduced from the speaker or the headphones has the sound image localized at a desired position after the crosstalk to both ears of the listener is canceled. Of course, when a speaker is used, the sound source can be localized at a desired position without the sound source being localized at a position closer than the intended position. In addition, when headphones are used, the sound is not caught around the ears, and the sound source is localized so that it is located at a distant position. In addition, there is no rise in the sound image, and it can be heard as an extremely natural sound, and it is possible to provide a virtual reality filled with a sense of reality. Further, in particular, according to the device of claim 3, since the added reflected sound and the direct sound are devised so as not to overlap with each other, interference between the reflected sound and the direct sound is avoided, and a sound image is formed. It has the effect of making it unclear.

[Brief description of drawings]

FIG. 1 is a schematic block diagram showing an embodiment of a sound image localization control device of the present invention.

FIG. 2 is a flowchart for obtaining a coefficient group.

FIG. 3 is a waveform diagram showing a process of adding a reflected sound.

FIG. 4 is a diagram showing a calculation example of a cancel filter.

FIG. 5 is a diagram summarizing experimental data when various reflected sounds are added.

FIG. 6 is a diagram showing a conventional example.

FIG. 7 is a block diagram showing an HRTF (head related transfer function) measurement system.

FIG. 8 is a waveform diagram of an impulse response measured in free space.

FIG. 9 is a diagram for explaining a localization change of a sound image with respect to temporally different emission sounds.

FIG. 10 is a diagram for explaining the quality of the sound image localization corresponding to the level difference of the emitted sounds that are temporally different.

[Explanation of symbols]

 1-3 Input terminal 4 Analog-digital converter 5 Serial-parallel converter 6-9 Convolver 10 ROM 11 Control sub CPU 11,12 Adder 13,14 Output terminal SP1, SP2 Speaker

Claims (3)

[Claims] 1. A pair of transducers separated from each other reproduces a signal processed by a pair of convolvers to which the same sound source is supplied, and a signal is reproduced to a listener at an arbitrary position different from the pair of transducers. In a sound image localization control device that makes the user feel that the sound image is localized, a pair of convolvers that perform convolution processing of signals from the same sound source according to a set coefficient, and a head measured at each sound image localization position. Based on the transfer function, a storage means for holding a cancellation filter coefficient group calculated as an impulse response, and a coefficient corresponding to a designated sound image localization position are supplied from the storage means to the pair of convolvers. And a canceling filter coefficient group for generating a reflected sound that precedes the direct sound coming directly to the listener and comes later than the direct sound. A sound image localization control device comprising a coefficient group calculated based on an added virtual impulse response. 2. The lead time of the reflected sound with respect to the direct sound is 9.1 msec to 11.nsec based on the maximum level of these sounds.
The sound image localization control apparatus according to claim 1, wherein the maximum level of the reflected sound is set to -10 dB to -13 dB with respect to the maximum level of the direct sound while being set to 3 msec. 3. The sound image localization control apparatus according to claim 2, wherein the reflected sound added prior to the direct sound is converged to zero before the direct sound starts.