JPH09244663A - Transient response signal generating method, and method and device for sound reproduction - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a transient response signal generating method which is simple and reproduces a sound while maintaining basic sound characteristics in a space and further a method and device for sound reproduction which use a transient response signal generated by the method. SOLUTION: Only pulses of a specific transient response signal are extracted from transient response signals used to reproduce a sound to generate a new transient response signal. The pulses may be extracted in the order of the absolute values of amplitude peaks on the basis of an optionally set threshold value or selected out of specific groups, negative pulses following extracted plus pulses may be extracted, or the extraction may be limited to only a transient response signal in a specific period. Further, the delay time of each pulse of the new transient response signal generated by the transient response signal generating method and relative amplitude values are set, a sequentially inputted sound signal is outputted at every delay time according to each relative amplitude value, and signals which are outputted at the same time are added to output a new sound signal.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating a new transient response signal extracted from a transient response signal used to reproduce sound that is generated and propagates in a three-dimensional space, and a transient generated by the method. The present invention relates to a sound reproducing method and device by a response signal.

[0002]

2. Description of the Related Art Regarding a sound reproduction method in a three-dimensional space,
A lot of research has been done so far. For example, the ray method and virtual image method based on classical geometric acoustics are the most common methods at present. However, these methods are not sufficiently realistic in sound reproduction because the wave precision is neglected and the calculation accuracy is extremely low because phase information is not entered. On the other hand, there are finite element method, boundary element method, etc. as the calculation method considering the wave nature.
Since a large amount of calculation is required to include a band including a high frequency such as Hz in the calculation result, it is extremely difficult or impossible to obtain the calculation result even with the currently developed supercomputer. However, it took a long time and was not practical.

On the other hand, the present inventor has proposed a processing method and apparatus for reproducing acoustic characteristics such as wave nature and phase together with a basic theoretical formula which is a modification of the Kirchhoff integral equation known as an integral display form of three-dimensional wave equation. Application number flat 7-7
3658 and the drawings.

This processing method and apparatus uses a transient signal having a sharp rising edge such as a delta approximation function as a sound source, and this transient signal propagates in a wave in space.
The present invention is a processing method and apparatus for obtaining acoustic characteristics that reflect on a boundary that divides a dimensional space and affect a predetermined observation point as a transient response signal, and that can obtain such a transient response signal efficiently and easily.

The advantage of processing with transient signals here is that when a continuous or discontinuous sound generation is divided into very short time intervals, each divided signal is treated just as a transient signal. Therefore, if the relationship between the response signal that spreads on the time axis in space and affects the observation point, that is, the transient response signal and the transient signal, becomes clear,
It is to be able to reproduce sound by applying it to a series of sounds.

[0006] Here, reproducing the transient response signal in the entire audible range and very realistically reproducing the frequency characteristic and the phase characteristic means obtaining the accuracy of the transient response signal in very detail. . However, this is completely contrary to the purpose of reproducing realistic sound with a simple and inexpensive device.

That is, in order to faithfully digitally reproduce a transient response signal including a large number of pulses having a frequency modulation, a phase delay, etc., a transient signal of a very short time propagates in various spaces, It is necessary to find and store the transient response signal at a very fine time interval (so-called sampling interval, the reciprocal of which is the sampling frequency) until the transient response signal decays and converges and can be ignored. Furthermore, in order to reproduce a continuous music or the like by the stored transient response signal, all the data of such a very large number of transient response signals must be constantly added in every sampling interval in real time.

For example, if 48 KHz digital sampling is performed to reproduce a frequency of 20 KHz, and the reverberation time to be reproduced is 1.3 seconds, 65,536 sampling numbers are required. Two
It is necessary to add 65,536 pieces of transient response data for every 0.8 us (the reciprocal of 48K) to perform analog output. Such a high-speed product-sum calculation device is very expensive and not practical.

[0009]

Therefore, the present invention utilizes such a highly accurate transient response signal to provide a new transient response for sound reproduction in a simple method while maintaining the basic acoustic characteristics in space. The present invention provides a signal generation method, and a sound reproduction method and apparatus using a transient response signal generated by the method.

[0010]

Therefore, according to the present invention, a new transient response signal is generated by extracting only a predetermined transient response signal from the transient response signals for the transient signal used for reproducing sound. And a transient response signal generating method. The extraction is performed by selecting, from the transient response signal, in order of magnitude of the absolute value of the amplitude peak, and also by setting a threshold value corresponding to the delay time at which each pulse of the transient response signal occurs, and corresponding to the time. Also, by selecting a pulse whose amplitude peak absolute value of the transient response signal exceeds the set threshold value, the transient response signal is divided into a predetermined group, and at least one pulse is selected for each of the predetermined groups. Can also be done by. The extraction is further performed by selecting a negative pulse following the extracted positive pulse, and may be limited only to a transient response signal within a predetermined period.

Furthermore, the present invention relates to a sound reproducing method and device for performing sound reproduction based on a new transient response signal generated by any one of the above transient response signal generating methods. More specifically, the delay time of each pulse of the new transient response signal generated by the transient response signal generation method and the relative amplitude value are set, and the acoustic signals sequentially input are set at the delay time intervals. The present invention relates to a sound reproduction method and apparatus for outputting a new sound signal by adding the output signals at the same time and outputting the sound signals according to the relative amplitude values. Further, the present invention relates to a filter for attenuating a predetermined frequency component with respect to the new acoustic signal, or a sound reproducing method and device for adding reverberation to the new acoustic signal.

[0012]

The present invention can be applied to any transient response signal for any acoustic reproduction, but in order to facilitate understanding of the present invention, application number 7-73658 by the present inventor is used. The transient response signal generation of the present invention will be described with reference to the disclosed means.

First, the Kirchhoff integration and the approximate boundary integration method will be described with reference to FIG. The Kirchhoff integral equation (Equation 2) is an integral equation (Equation 2) obtained by solving the three-dimensional wave equation (Equation 1), and the modified integral equation (Equation 3) is called an approximate boundary integral equation.

[0014]

[Equation 1] Three-dimensional wave equation Where φ is velocity potential, c is sound velocity, t is time,
F is a wave source, x, y, and z are three-dimensional coordinates, S2 in FIG. 14 is a surface forming an arbitrary space, Q is an arbitrary point on S2, P is an observation point for obtaining a transient response, and V is a point near the observation point P. area excluding the area V1, ds ₂ is a differential surface vector on S2.

[0015]

[Equation 2] This is an integral form of Kirchhoff's three-dimensional wave equation, and represents the velocity potential at the observation point P. Here, φp is the velocity potential at the observation point, []
td is the delay time tr / c, V is an arbitrary space surrounding the observation point P, S2 is the surface of V, Q is an arbitrary point on S2, and r is the distance between Q and the observation point P.

[0016]

[Formula 3] Approximate boundary integral equation obtained by modifying Kirchhoff's integral equation Here, φp is the velocity potential at the observation point, []
Any space td is the delay time t-r / c, V is surrounding the observation point P, the surface of the V S2, Q is any point on S2, r is Q and the distance of the observation point P, r ₀ is Q from the wave source The total propagation distance up to is shown. Further, f (t) is a transient signal of a sound generated from the wave source.

Next, the means for generating the transient response signal based on the approximate boundary integral equation of Equation 3 will be described. still,
The means described here is not limited to this approximate boundary integral equation, and any means can be applied as long as it determines the wave characteristics.

FIGS. 1, 2, 3 and 4 show concrete flow charts for obtaining a transient response signal (impulse response signal).

FIG. 1 shows the procedure of the entire process, and as shown in the figure, it can be roughly divided into two, process A and process B. "Process C (process 7)" in FIG.
2 is shown in FIG. 2, and the processing procedure corresponding to “Process D (Process 10)” in FIG. 2 is shown in FIG. FIG. 4 shows a processing procedure corresponding to the processing E shown in FIG.

In process 1 [initial setting] which is the first process of FIG. 1, "calculation condition setting", "wave source condition setting" and "boundary setting" which are prerequisites for performing the process.
And "setting of source radiation vector".

"Setting of calculation conditions" means the three-dimensional coordinates of the wave source, the number of observation points, the three-dimensional coordinates of each observation point, the temperature and humidity of the air in the space, the analysis frequency (the highest included in the transient response to be calculated. Frequency), the duration T of the transient response to be calculated, etc. are set. The setting is performed by direct input from an input device or by reading from an external storage device.

In the "setting of wave source condition", a transient signal which is an initial value of the sound from the wave source is given by a delta approximation function close to an impulse and its derivative, but this initial value may be changed according to the purpose. .

Further, the propagation velocity is calculated from the air temperature and the humidity indicated by the initial setting. From the analysis frequency, a discrete time interval corresponding to a frequency twice or more the analysis frequency is set according to Shannon's sampling theorem. The size of the storage area for storing the calculated transient response used in the process B is determined from the duration of the transient response and the discrete time interval.

In the "boundary setting", information about the boundary that constitutes the field is set. This is a numerical representation of a space that simulates wave propagation, and is constructed by a large number of polygons (planes). Here, each polygon is called a boundary, and the number of boundaries, the normal vector of each boundary, the coordinates of each vertex of the boundary, the reflection and absorption of the boundary, and the like are set. In the following description of the specific example, in order to facilitate understanding of the present invention, the boundary is only total reflection or total absorption in all frequency bands.

In the "setting of wave source radiation sound ray vector", the number N of sound ray vectors radiated from the wave source is calculated. This is because a wave propagating from a wave source is simulated, a location and a time at which the wave is reflected at a boundary are calculated, and a velocity potential φ which is an acoustic characteristic of an observation point is calculated by using an integral equation of Expression 3.
_{This is} for calculating _p . The simulation of the wave propagation at this time is performed by using an infinite number of vectors radiating from a wave source into space at an equal solid angle. This is defined as a sound ray vector. The number N of radiating sound ray vectors is the duration T from the wave propagation start.
Is passed, the distance between the adjacent sound ray vectors is within 1/2 of the wavelength λ of the analysis frequency, preferably 1/4.
Set to within. However, depending on the performance of the processing device or the degree of approximation for reproduction, the distance between adjacent sound ray vectors may be λ equal to or more than １ ／. Note that vectors indicating a wavefront traveling in space are collectively referred to as a sound ray vector, and particularly a sound ray vector radiated from a wave source is referred to as a radiation sound ray vector.

In process 3, the wave propagating from the wave source is simulated by a sound ray vector. First, N
Direction vector Dn of the n-th radiated sound ray vector so that the directions of the radiated vectors are the same solid angle from the wave source.
Is calculated.

Further, in process 4, the boundary B having an intersection with the sound ray vector traveling from this wave source is selected by calculation.
At this time, if there is no boundary having the intersection with the sound ray vector, calculation is performed for the next radiated sound ray vector (process 5).

When there is an intersection with a certain boundary, the sound ray vector is reflected at the boundary B, but it is necessary to judge whether the boundary is the front or back of the sound ray vector.

Whether the surface is the front side or the back side is determined by determining that the angle between the normal vector of the boundary and the sound ray vector is 0 ° to 180.
° or not. That is, if the normal vector of the boundary is defined so as to face the outside of the virtual space, the mutual angle between the sound ray vector and the radiation vector is 0 °.
In the case of １８０180 ° (that is, when the inner product is positive), the boundary is defined as a table, and in the case of 180 ° to 360 °, the boundary is defined as a back. This is because the ray vector is from the inside of the boundary (ie, the front side of the boundary) and the sound is reflected at the boundary, or the ray vector is from the outside of the boundary (ie, behind the boundary). Therefore, it is possible to determine in computer processing whether a boundary exists only in the sound ray vector direction and does not become a boundary contributing to sound reflection.

According to the above definition, in the process 6, when the boundary is the back side of the sound ray vector, one having an intersection at another boundary is selected. If the boundary B is in the table for the sound ray vector, the process proceeds to processing C.

In the process C shown in FIG. 2, first, if the boundary B is set in the initial setting so as to completely absorb the wave, the process C ends (process 11). Otherwise, the propagation distance d from the wave source to the intersection Q of the boundary B is calculated (process 12). When the wave propagates along the propagation distance d, if the distance exceeds the duration T, processing 1
Ends. Processes 11, 12 and 13 shown in FIG.
Even if the order of the processing is different from that of FIG. 2, a series of expected processing results can be obtained. Therefore, any order may be used, but the order shown in FIG. 2 is the most efficient in terms of processing time.

Next, in process 14, the incident angle α of the sound ray vector incident on the boundary B is calculated. This is for obtaining the incident angles of the radiation and the sound ray vector of the boundary plane in Expression 3.
The incident sound ray vector is the n-th radiated sound ray vector Dn
The sound ray vector arriving at the boundary B is denoted by En, r. Therefore, En, 1 is equal to the radiation sound ray vector reaching the boundary B from the wave source.

Next, the direction vector of the reflected sound ray vector at the intersection Q of the incident sound ray vector is calculated. The reflected sound ray vector refers to a sound ray vector at the boundary B where the n-th radiated sound ray vector Dn is reflected at the boundary B, and the reflected sound ray vector reflected at the r-th boundary is Fn, r.
Therefore, Fn, 1 is a sound ray vector in which the n-th sound ray vector is reflected at the boundary for the first time.

In process 16, the reference number B of the boundary B, the total propagation distance d from the wave source to the intersection Q, the three-dimensional coordinates of the intersection Q,
The direction vector En, r of the incident sound ray vector and the direction vector Fn, r of the reflected sound ray vector are stored in a memory or an external storage device. A series of storage of the data calculated each time the radiation sound ray vector Dn is inverted at the boundary is called a propagation history of the radiation sound ray vector Dn. The reason for storing the propagation history is that, in the process D of FIG. 3, every time the sound ray vector is reflected at an arbitrary point (here, each boundary) on S ₂ shown in Expression 3, the velocity potential contributing to the observation point P is calculated. It is to calculate. Here, the total propagation distance d corresponds to r ₀ in Equation 3.

Next, another boundary B having an intersection with the reflected sound ray vector having the direction Fn, r is selected, and the intersection at that time is newly designated as Q. If there is no boundary B having an intersection, the process ends (step 18). If there is an intersection, but the incident sound ray vector is incident on the boundary B from behind, a search is made for an object having the intersection Q from another boundary. If it is incident on the surface, the process is continued (process 19).

Process C for the n-th radiation vector
Is completed, as shown in FIG. 1, the calculation is repeated for the next radiated sound ray vector. When the processing is completed for all the radiated sound ray vectors, the processing B is performed.

Therefore, the process A showing a series of processes 1 to 8 in FIG. 1 is performed in order to calculate and store the propagation history of each sound ray vector.

From the above, the apparatus for achieving the above processing A is
It has means for storing initial settings, processing means for calculating propagation record data, and means for storing propagation record data.

In process D, as shown in FIG. 3, calculation is performed based on the propagation history of the n radiation sound ray vectors calculated in process 1.

First, in process 21, the coordinates of the first observation point P for which the transient response is calculated are read from the memory or the like. Then, first, data relating to the propagation history of the radiation sound ray vector of n = 0 is read from the storage (processing 22). When only the back side of the boundary is seen from the observation point P with respect to the first boundary B of the read propagation history, the next successive propagation history is read (process 32) and the process is continued. On the other hand, if the boundary B is a surface when viewed from the observation point P, a direction vector R from the observation point P to the intersection Q described in the propagation history is calculated (processing 25). The distance RD is calculated (process 26). If the straight line connecting the observation point P and the intersection point Q intersects another boundary, it is determined that the velocity potential from the intersection point Q does not affect the observation point P, and the next successive propagation history (processing 32) Continue processing for. When the straight line connecting the observation point P and the intersection point Q does not intersect with another boundary,
It is determined whether or not the wave arrives at the observation point P within the time T (processing 28). If it does, processing E is performed on the assumption that the velocity potential on the boundary B affects the observation point P.
Whether the wave arrives at the observation point P within the time T is determined by determining the distance obtained by adding the total propagation distance d to the intersection Q and the distance RD within the duration T of the initially set transient response. It is determined by whether or not it propagates.

As shown in FIG. 4, the processing E calculates the velocity potential created at the observation point P by the minute surface element on the boundary represented by the intersection Q based on the equation 3, and determines the time at which the observation point P is affected. Based on that, the transient response is stored and stored in the array. At this time, if a value already exists at the storage position, the value is added. Specifically, it is as follows. Here, the minute plane element is defined as an intersection Q by a solid angle when a sound ray vector is defined.
A surface made in FIG. 5A shows a plurality of sound ray vectors 50 directed to a boundary 52, and shows a small plane element 54 on the boundary 52 formed by one sound ray vector shown by a solid line. FIG. The micro surface element 54 on the boundary when the boundary is viewed from the side is shown. As is clear from the figure, the area of the microplane element strictly depends on the angle between the sound ray vector and the boundary, but in the processing, the total propagation distance d to the intersection Q and the bottom of the cone formed by the solid angle are used. The area 56 may be used. In this case, although the accuracy of the approximation value is somewhat degraded, it can withstand practical use, and the processing speed is improved because the minute area is determined only by the distance and the solid angle. In addition, this small surface element 5
The area of 4 corresponds to ds ₂ in Equation 3.

Returning to FIG. 4, the process 41 is for calculating the first term in the integral term of the equation 3, and the process 42 is for the equation 3
Is to calculate the second term in the integral term. In this embodiment, since the reflection and absorption of the sound at the boundary are total reflection or total absorption, the function f (t) can use the transient function of the wave source set in the initial setting. If the reflection and absorption are partial reflection and absorption, f (t) corresponding to the characteristics of the boundary may be determined for Fn, r of the propagation history every time the light is reflected at the boundary.

The process 43 is for obtaining the area of the above-mentioned minute surface element in order to obtain the integral approximation value.

In process 44, the product of the result calculated in process 41 and the area of the minute surface element calculated in process 43 is calculated, and the initial value of the wave source is convoluted.

In process 45, the result calculated in process 42,
The product of the area of the small plane element obtained in the process 43 is obtained, and convolved with the differential value of the wave source initial value.

Processes 46 to 48 are methods for storing the transient response from the approximate boundary integration result. In order to reproduce the acoustic characteristics at the observation point P, a numerical array in which the calculated transient response is stored is provided. This array may be provided in the initial setting in the process A. Since the arrangement position j at which the transient response is stored is determined in accordance with the time Dt at which the wave arrives at the observation point P, the result of the transient response according to the time at the observation point P is converted into a corresponding numerical array. It is added and stored.

First, in process 46, the time Dt when the wave arrives at the observation point P is obtained from the sum of the total propagation distance d to the intersection Q of the sound ray vector and the distance RD.

Next, in process 47, the position j of the numerical array corresponding to the arrival time Dt is obtained. From the above,
In process 48, the time-series data obtained in processes 44 and 45 are added to the corresponding array position j of the numerical array and stored. "Adding and storing" means that when a value is already stored at the array position j, the value is added to that value and stored, whereby the same effect as that obtained by the approximate integration can be obtained. Further, by arranging the time series, it is possible to read out the sound field characteristics at the observation points P for each arrangement order and reproduce the sound, thereby achieving an efficient sound reproduction method and means. it can.

When the process E is completed, the process is continued for the propagation history of the next radiation ray vector. When the calculation is completed for all the data of the propagation history for the nth radiation ray vector, the n + 1th radiation ray vector is calculated. Similar processing is performed for the vector. Then, when the calculation is completed for all the radiation sound ray vectors, the processing is performed for the next observation point as shown in FIG.

Therefore, in the process D, from all the propagation histories of each sound ray vector, the velocity potentials reflected by the sound ray vector at the boundary and contributing to the observation point P are integrated, and the observation point P is calculated.
Is obtained and stored.

From the above, the means for achieving the processing D may be any means for applying the stored propagation history data to the approximate boundary integration method, and an arithmetic processing unit for performing arithmetic operations in accordance with a series of steps of software or It may be a computer, or may be a device in which the processing procedure of the approximate boundary integration method is configured in advance as hardware.

When the process D is completed, the process 49 of FIG.
As shown by, in order to represent the acoustic characteristic given by the direct sound of the sound source to the observation point P, the transient characteristic is added to the arrangement of the arrangement position j corresponding to the time when the direct sound arrives at the observation point P and stored.

In FIG. 6, the potential applied to the observation point P is calculated for each boundary at which each sound ray vector is reflected when the processing B is performed, and the potentials are added by time and stored, and the transient response at the observation point P is obtained as a whole. The procedure is shown in the form of a graph, and the process D and the addition of the direct sound will be obvious.

FIG. 7A shows a response signal obtained by the virtual image method which is one of the conventional methods, and FIG. 7B shows the amplitude waveform of the transient response signal actually obtained by the above processing. Is. The amplitude waveform of FIG.
It is clear from comparison with (a) that the negative waveform is reproduced in addition to the positive waveform. As described above, it has become possible to reproduce a transient response amplitude waveform which has conventionally been difficult to obtain or which requires an enormous processing time to obtain.

Next, a method of reproducing an actual sound according to this waveform will be described. FIG. 8 shows an analog sound signal reproduced by a sound reproducing device 70 such as a CD player, a record, or a cassette, which is input from a line 71 to a product-sum operation device (also referred to as a convolution device) 72, A diagram is shown in which sound is reproduced from a speaker 82 through an amplifier 80 and output as an analog output to 79. The convolution device 72 includes a central processing unit 76 that actually performs a product-sum calculation (convolution), and an A / D converter 73 and a D / A converter 77 connected before and after the central processing unit 76 by lines 75 and 77, respectively. However, when the sound reproducing device 70 has a digital output, the A / D converter 73 is naturally unnecessary. Such a product-sum operation device is, for example, a Lake DSP Pt
y. Digital Audio Convol commercialized by Ltd
ution Processor FDP 1 plus is available.

Next, referring to FIG. 9, the concept of a series of processes in which the digital signal from the A / D converter 73 is input to the central processing unit 76 and the processed digital signal is output to the line 77 will be described. .

The discrete signal 80 of FIG. 9 corresponds to the line 71 of FIG.
The analog signal appearing above is shown discretely with respect to time. In practice, the time and amplitude data of this waveform are input to the central processing unit 76 as a digital signal.
The central processing unit 76 outputs the transient response signal 83 obtained by the above-described transient response signal generation processing or by another method.
Is stored. In order to determine the discrete digital signal 86 to be actually determined, a response signal corresponding to the transient response signal 76 is generated for all of the divided discrete time data of the discrete signal 80 input in real time. The processing is performed by the convolution unit 84. Since it is necessary to reproduce all the sound generated with the delay of the transient response signal, this processing is an operation of adding the number of time divisions of the transient response signal at each discrete time interval. For example, 20KHz
48, per second to reproduce up to the frequency component of
Using a transient response signal that lasts for 1.3 seconds divided by 000 times requires 65,536 taps, so that 20.8 us (1 / 48,0
00), data of 65,536 taps must be added to perform analog output.

This processing is performed by the Lake DSP described above.
Pty. Digital Audio Convolution Pr by Ltd
Ocessor FDP 1 plus is suitable, but on the other hand, it is necessary to reproduce the sound in a simpler and cheaper method. Therefore, embodiments of the method and apparatus will be described below.

FIG. 10 shows the logarithmic representation of the amplitude (FIG. 10A) and energy of the transient response signal generated by the same method as the transient response signal generation method of FIG. 7B.
(B) shows a time change. The use discrete time interval of the transient response signal is set to 2 corresponding to 48 kHz sampling.
By setting the reproduction duration to 1.3 seconds and 0.8 μs, the total number of samples of the transient response signal on the time axis was set to 65,536. Here, each number from 0 to 65,536 of the sampling number is defined as a sampling number.

FIG. 10A shows a transient response signal waveform that appears after the sampling number 0 is set to the time when the transient signal as the sound source is generated. As can be seen, the major pulses are concentrated in the first half up to about 4,000 sampling numbers. FIG.
From (b), the decay of the transient response is represented by sampling number 6
The attenuation of about 60 dB is shown at 5,536, and it depends on the size of the space to be used and the wall material. However, in this case, it is sufficient to set the duration to 1.3 seconds. .

FIG. 11 is an enlarged view of the waveform of FIG. 10A from 0 to 4,096 sampling numbers.
It shows with (a). This reveals that the portion having a particularly large amplitude is further limited even up to about 4,000 sampling numbers where the main pulse exists. From this result, another new transient response signal (FIG. 11) in which only the characteristic pulse of FIG.
(B)) was prepared. Further, the extracted transient response signal is shown in FIG.
When the sound was reproduced by using it as the transient response signal 83, it became clear that the spatial spread of the sound could be recognized.

From the above, a pulse having a large amplitude, which reaches the human ear intermittently over time, is like an exponentially decaying echo that has conventionally been used to represent spatial reverberation. Unlike reverberation, it can stimulate human hearing and cause humans to recognize the three-dimensional spatial characteristics and unique expanse. Therefore, it is considered that by reproducing this characteristic portion, it is possible to reproduce the main characteristics that can be heard at a predetermined observation point (listening point) in the space of the sound generated in the space to be reproduced. On the other hand, the other parts of the pulse, for human hearing,
Since it can be considered as a complementary element of the above-mentioned characteristic pulse, it is considered that the characteristic pulse portion need not be reproduced faithfully.

Further, it is clear that the real image of the space can be recognized by using the transient response signal containing the negative pulse shown in FIG. 11B together with the positive pulse appearing before it. became. The effect of introducing a negative pulse means that the frequency component of the transient response can be extracted, and at the same time, the human being can recognize the three-dimensional shape of the space, and particularly, the discontinuous boundary of the space. . This is because, as can be logically derived from the mathematical formulas (2) and (3) above, a negative pulse means that the space appears due to a discontinuity in the space. is there. That is, this negative pulse is generated by the side of the part where the wall surfaces (boundaries) forming the three-dimensional space intersect with a convex or the part where the material is greatly changed. It is thought that it appeared as a real image of. Therefore, reproducing a negative pulse together with a positive pulse is important in reproducing sound in a three-dimensional space.

In this way, if the pulse of the characteristic portion of the transient response signal is extracted and the characteristic portion of the sound is reproduced by using the extracted transient response signal, the acoustic characteristic in the basic space is reduced. Can be recognized. Furthermore, by limiting and extracting data to be a new transient response signal from a large amount of transient response signal data, the number of transient response signal data to be added is dramatically reduced, and The burden can be reduced.

As a method of extracting a characteristic portion of the transient response signal, (1) a method of extracting in order of the maximum value (absolute value) of the pulse amplitude is simple. this is,
It is an audible stimulus exerted on the human ear as a pulse amplitude. When extracting like this,
In order to avoid an unlimited increase in the number of pulses to be extracted, the number of pulses may be limited numerically, or a certain threshold value may be set and only a pulse amplitude larger than that may be extracted. (2) A certain threshold value is set, and only pulse amplitudes (absolute values) that are equal to or larger than the threshold value are selected in order from the origin on the time axis (generally, when a sound source is generated). This is effective when the beginning of sound source generation is particularly important, and pulses below that threshold are not particularly useful.
(3) In the above 2, when the predetermined number of selections is reached, the extraction is terminated. This is particularly effective when there is a limit on the calculation processing capacity such as the number to be added. (4) For each signal group considered as a group of signals, select the pulse having the strongest pulse amplitude or the first pulse from the group. Taking FIG. 12A as an example, the signal group 90 can be regarded as a group of signals unlike other signals. As shown in FIG. 12B in which the transient response signal is temporally enlarged, discretely scattered peaks can be seen from about 1024 to 2,000 sampling numbers. In some cases, the time interval between the peaks may be a short time equal to or less than the human hearing resolution. In this way, it is not necessary to reproduce (ie, extract) all the peak pulses scattered at a time interval less than the human hearing resolution, and it is possible to sufficiently appeal to the human hearing just by extracting one of them. It is. For example, as is known as the Haas effect, signals that are continuous at short time intervals such as 10 ms cannot be distinguished. The one pulse selected in this case could be the first peak pulse in a group or the largest peak pulse in the group. FIG. 12C is a diagram in which only the first pulse of the pulse group 90 is extracted based on this means.
(5) A fixed period is set for each of the means,
The above means may be applied only during that period.
In the transient response signal in space, the amplitude of the sound wave exerted on the observation point is determined by the overlap of the direct wave or the reflected wave of the wave source spreading in various directions, but after a predetermined time, the energy of the wave source is attenuated and sufficient Cannot produce peak value. For example, as is clear from FIG.
At the time of the 0 sampling number, the energy amount of the observation point shows an attenuation of about 60 dB, and it is not necessary to perform the extraction over the period. Although the period is divided by the amount of energy attenuation, it can be divided by the amplitude value. Referring to FIGS. 10A and 11A, the region where the amplitude value is 0.005 or more is 4096.
It is up to the sampling number, and does not exist thereafter.
Therefore, a peak pulse may be selected up to that period. (6) In each of the extracting means, it is also effective to extract a negative peak pulse which is paired with the positive peak pulse and follows the pulse. As mentioned earlier, the negative peak pulse combines with the positive peak,
This is because not only the frequency component can be reproduced, but also a real image of the space can be grasped three-dimensionally.

As described above, some pulse extraction means have been presented, but without using a high-performance product-sum calculation device,
An apparatus for reproducing sound based on the transient response signal after the extraction by a simple method will be described with reference to a conceptual diagram of FIG. FIG. 13 shows a simple convolution device 100 instead of the central processing unit 76 of the convolution device 72 in FIG. The simple convolution device 100 includes a multi-tap delay 110, a reverb unit 120, and an equalizer unit 130.

The multi-tap delay 110 includes a delay line 112, a weighting section 114, and an adding section 1.
The adder 118 has a plurality of output lines 115 for tap signals from the delay line 112 via the weighting unit 114.

The multi-tap delay 110 processes the input acoustic signal based on the delay time and the amplitude set in advance, and outputs the signal.
The function of the multi-tap delay 110 will be described using the transient response signal obtained by extracting the transient response signal of FIG. Signals to be extracted include P1 to P30 in FIG.
Up to 30 pulses.

First, a delay time corresponding to the pulse position of P1 is set in the delay line 112 for the first tap output of the plurality of tap outputs 115. Here, P1 represents a direct sound reaching the observation point, and its time is 16.52 ms (7 as a sampling number).
It is the 93rd, and its occurrence time is 16.52 ms = 79
3x1 / 48,000)). The delay line 112 holding the set value can output the signal input from the line 75 to the first weighting unit 114 after a delay time of the set value.

Further, the weighting unit 114 sets the size of the signal to be reproduced. In this case, it is set relatively as each ratio of the magnitude of each pulse amplitude of the transient response signal to be reproduced. In this example, the pulse having the maximum amplitude value is P29.
The set value of 1 is about 70% (0.7). Therefore, 70
The first weighting section 114, which has set and held%, weights the amplitude of the signal obtained from the delay line 112 by 70% (generally multiplies by 0.7 which is a ratio of the set value), and sets the first tap. It is possible to output to the output 115.

With the above setting, the signal input from the line 75 becomes 70% of the amplitude of the input signal from the first tap output of the plurality of tap outputs 115 for the pulse of P1 after 16.52 ms has elapsed. The sound is output as described above, and can reach the listener's ears as a direct sound after 16.52 ms have elapsed since the sound source was generated.

Next, the setting for P2 is set to the next tap output 11
5 is performed. The delay value to be set is 22.16
ms (= 1064 / 48,000), and the amplitude setting value is 63%. With this setting, the first characteristic sound of the reflected sound can reach the listener's ear 22.16 ms after the sound source is generated.

In the same procedure, up to P30 is set, but with respect to the negative pulse, it is set that the pulse is a negative pulse when the amplitude setting value is set. When all the settings are completed in this manner, the multi-tap delay 110 outputs each tap output 1
The signal is sequentially output to each tap output 115 in accordance with the delay value set for 15.

The signal input to the line 75 is generally a continuous acoustic signal, and the acoustic signal input in the past has a predetermined tap output 115 after a predetermined delay time.
And at the same time, the audio signal input thereafter is output from another predetermined tap output 115 after a predetermined delay time, so that all the signals output to each tap output 115 at the same time are added. For this purpose, an adder 118 is provided. Therefore, all the signals at the same time for P1 to P30 are added at one sampling interval by the adding section 118 (1 when the sampling frequency of the input signal is 48 KHz).
/ 48,000 seconds), and the result is sent to the output 116.

Next, it is desirable to add reverberation by the reverb unit 120 so that the listener can feel the actual sound field using the signal output to the line 116. The reverb unit may be a commercially available reverberation device, and generally adds sound so that the signal generated at the multi-tap delay 110 for each sampling frequency attenuates exponentially with time. In this case, the reverberation time and the characteristic of attenuating the reverberation may be arbitrary depending on the reverberation device used. By adding the reverberation, the gap between the pulses of the newly extracted transient response signal can be easily filled. At the same time, the sense of sound caused by the extracted characteristic pulse is not lost. In particular, the addition of reverberation is particularly effective when one pulse is selected from the group of signals (5) shown by the extraction means.

The reverberated signal is input to the equalizer section 130 via the line 117. Line 116,
The signal appearing on the signal 117 is obtained by extracting only a pulse-like characteristic signal from the transient response signal by the extraction method according to the present invention, and therefore contains a very large number of high frequency components. . The equalizer unit 130 is provided to remove the high frequency component, functions as a low-pass filter, and has a role of changing the frequency characteristic of the output to the line 77.

As described above, according to the present invention, only a characteristic pulse is selected and extracted from a large number of transient response signals to generate a new simple transient response signal.
A very simple product-sum calculation device, or a tap with a limited number of taps, is used because it is possible to perform acoustic reproduction in a three-dimensional space by performing a product-sum operation on only a limited number of transient response signals.・ It became possible to use a delay device, and it became possible to easily reproduce the sound in a virtually constructed three-dimensional space.

[Brief description of drawings]

FIG. 1 is a flowchart for obtaining a transient response signal.

FIG. 2 is a flowchart showing processing for obtaining and storing a propagation history.

FIG. 3 is a flowchart of a process for calculating a transient response from the propagation history of each sound ray vector by an approximate boundary integration method.

FIG. 4 is a flowchart of a process of calculating a transient response by an approximate boundary integration method and adding and storing the transient response.

FIG. 5 shows a small surface element created by a sound ray vector and its area.

FIG. 6 is a transient response obtained by adding the potential brought to the observation point each time each ray vector is reflected, by adding the entire ray vector and the direct sound along the time axis.

FIG. 7 is a comparison between the impulse response of the virtual space obtained by the method of FIG. 1 and the response by the virtual image method which is a classic calculation method in the related art;

FIG. 8 is a block diagram of an apparatus for reproducing sound using a transient response signal.

FIG. 9 is a conceptual diagram for reproducing sound using a transient response signal.

FIG. 10 is a graph showing the amplitude and energy decay of the obtained transient response signal.

FIG. 11 is a graph enlarged on a time axis showing the obtained transient response signal and a characteristic transient response signal.

FIG. 12 is a graph enlarged on a time axis showing another transient response signal similar to FIG. 11 and a characteristic transient response signal.

FIG. 13 is a block diagram showing a simple apparatus for reproducing sound using an extracted transient response signal.

FIG. 14 is a diagram illustrating a potential exerted on an observation point in space.

[Explanation of symbols]

50. . . Sound ray vector, 52. . . Border, 5
4. . . Micro surface element 56. . . Micro area, 70. . . Sound reproduction device, 7
2. . . Folding device 73. . . A / D converter, 76. . . Central processing unit 78. . . D / A converter, 80. . . Amplifier 82. . . Speaker 83. . . 84. transient response signal; . . Convolution part 90. . . Pulse group 100. . . Simple folding device 110. . . Multi-tap delay, 112. . .
Delay line 114. . . Weighting section 115. . . Tap output, 118. . . Adder 120. . . Reverb part, 130. . . Equalizer section

Claims (14)

[Claims] 1. A transient response signal generation method for generating a new transient response signal by extracting only a predetermined transient response signal from among transient response signals for a transient signal used for reproducing sound. 2. The transient response signal generation method according to claim 1, wherein the extraction is performed by selecting the amplitude of the transient response signal that varies on the time axis in the order of magnitude of the absolute value of the amplitude peak. 3. The extraction is performed by selecting, from the pulses of the transient response signal changing on the time axis, in the order of positive amplitude peaks, and also selecting a negative pulse following the positive pulse. The transient response signal generation method according to claim 1, further comprising: 4. The extraction is performed by setting a threshold value corresponding to a delay time generated by each transient response signal, and selecting a pulse whose amplitude peak absolute value of the transient response signal exceeds the set threshold value. 1. The transient response generation method described in 1. 5. The transient response signal generation method according to claim 1, wherein in the extraction, the transient response signal is divided into predetermined groups, and at least one pulse is selected for each of the predetermined groups. 6. The extraction is performed by dividing the transient response signal into a predetermined group, selecting at least one pulse for each of the predetermined groups, and also selecting a negative pulse following the selected pulse. The transient response signal generating method according to claim 1. 7. The transient response signal generation method according to claim 1, wherein the extraction is performed only on a transient response signal within a predetermined period. 8. A sound reproducing method for performing sound reproduction based on a new transient response signal generated by any one of the transient response signal generating methods according to claim 1. Description: 9. A delay time of each pulse of a new transient response signal generated by the transient response signal generating method according to claim 1 and a relative amplitude value are set, and are sequentially input. A sound reproduction method in which the sound signals are output in accordance with the relative amplitude values for each delay time, and the output signals are added at the same time to output a new sound signal. 10. The sound reproducing method according to claim 9, wherein the sound signal is passed through a filter that attenuates a predetermined frequency component with respect to the new sound signal. 11. The sound reproducing method according to claim 9, wherein reverberation is added to the new sound signal. 12. Storage means for storing a delay time of each pulse of a new transient response signal generated by the transient response signal generating method according to claim 1 and a relative amplitude value. Processing means for outputting sequentially input acoustic signals for each delay time according to the relative amplitude values, and addition for adding the output signals at the same time to output a new acoustic signal And a sound reproducing device comprising: 13. The sound reproducing device according to claim 12, further comprising a filter that attenuates a predetermined frequency component of the new sound signal. 14. The sound reproducing method according to claim 12, further comprising a reverb unit that adds reverberation to the new sound signal.