JP3226255B2 - Music synthesis system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music synthesis technique, and more particularly to a "physical-modeling synthesis" for synthesizing music in accordance with the mechanism of a natural musical instrument.
s), which refers to a musical sound synthesis technique known as "s).

[0002]

2. Description of the Related Art Today, tone synthesis based on physical models is
For example, "sampling" (or "waveform table")
It is commonly used along with current mainstream music synthesis methods such as synthesis and FM synthesis. Musical sound synthesis based on such a physical model is particularly useful for simulation of wind instruments and stringed instruments. By accurately simulating the physical phenomena of musical sound generation in natural musical instruments, electronic musical instruments can generate high-quality musical sounds.

[0003] In the case of a stringed musical instrument, a structure for synthesizing a musical tone typically includes a delay loop with a filter, that is, a closed loop for realizing a delay having a length corresponding to one cycle of a musical tone to be generated, and a closed loop. And a filter included in the filter. An excitation signal is input to the closed loop and circulates through the closed loop. Thus, the output signal of the closed loop can be extracted as a tone signal. This signal attenuates according to the characteristics of the filter. The filter also simulates attenuation at the strings and at the ends of the strings (eg, guitar nuts and bridges).

[0004] In a practical stringed instrument, the strings are acoustically coupled to a resonating body or resonating part, and the physical vibration of the strings excites the resonating part. Therefore, in order to accurately simulate a natural musical instrument, it is necessary to provide a filter on the output side of a delay loop with a filter. Also,
To obtain high quality musical sounds, it was necessary to imitate the string output with a large and expensive filter that simulated the instrument itself. Generally, the excitation signal is white noise or filtered white noise. Alternatively, the closed loop may be provided with a physically accurate "pluck" sound waveform as the excitation signal, thus more accurately simulating the string pluck sound.

[0005] The above-described conventional tone synthesis system is shown in FIG.
Is shown in The delay loop with a filter includes a delay element 10 and a low-pass filter 12. An excitation source (for example, an excitation table) 14 supplies an excitation signal to the delay loop via an adder 16. The contents of the excitation table can be automatically read from the memory table in response to, for example, a trigger signal generated in response to key depression. The excitation signal input to the filter-added delay loop circulates through the loop, and changes with time by the operation of the filter 12. Thus, a signal is extracted from the delay loop and supplied to the main body filter 18. High quality tone synthesis requires complex and expensive body filters (typically digital filters) or delay loops with additional filters.

[0006] One or more digital signal processing (D
It is more common to realize the tone generation by software using a chip for SP).
(1) can also be realized by hardware.

[0007]

The above-mentioned conventional tone synthesis system is capable of synthesizing an extremely high-quality tone, but requires a complicated and expensive main unit filter to simulate the instrument main unit. Problem.
SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide a physical model-type tone synthesis system that can easily synthesize high-quality tone at low cost.

[0008]

SUMMARY OF THE INVENTION A tone synthesis system according to the present invention is a tone synthesis system for synthesizing a tone generated by the cooperative action of a vibration element and a resonance member acoustically coupled to the vibration element. a is an input unit for receiving an excitation signal, a delay unit for delaying the signal, it will be a filter for filtering the signal and a closed loop connection, or
And an output unit for extracting an output from the closed loop.
Comprising a closed-loop means for delay in the closed loop corresponds to the pitch of a musical tone to be synthesized, and excitation means for supplying an excitation signal to the input portion having a component corresponding to the response of the resonant member It is characterized by the following.

[0009] The signal is supplied to the input of the closed loop means.
The excitation signal that is to be modeled according to the invention is
This is a signal having a component corresponding to the response of the vibration member. this
The response of the resonant member to be modeled
Closed loop means using a signal having a corresponding component as an excitation signal
Input, so the expensive body
Ruta can be eliminated.

According to one embodiment, the excitation means includes:
Having a component corresponding to a first partial response of the resonant member
Supplying an excitation signal to the input unit,
Further, the tone signal output from the output unit is
A resonance element for adding resonance according to the second partial response of the resonance member.
Vibration filtering means, said first partial response and
The overall response of the resonant member by a second partial response
And a tone having a characteristic according to this response is synthesized.
It is characterized by the following.

[0011]

In the tone synthesis system according to the present invention, no expensive main filter can be used at all, and the quality is the same as that of a tone generated by a conventional system including a complicated main filter. Can be realized.

In a system including a conventional delay loop with filter and a body filter as shown in FIG. 1, both the delay loop with filter and the body filter are linear, time-varying components. Therefore, it is possible to reverse the arrangement order of these elements. Therefore, an equivalent system can be realized by rearranging the main body filter before the delay loop with the filter (for example, FIG. 2). The output of the excitation signal generator in such a rearranged configuration is directly provided to the main body filter. The output signal of the main body filter, that is, the delay loop, is recognized by recognizing that when the string is plucked or hit, the excitation by the string generally takes the form of an impulse. It is determined that the excitation signal applied to the filter will represent the impulse response of the body filter.

[0014] In view of this, the <br/> Oite the embodiment described below, the impulse response is measured and the measured impulse response is stored as a set excitation signal.
In this way, the body filter can be removed,
The collective excitation signal is provided directly to the delay loop with the filter. Thus, by providing an appropriate excitation signal corresponding to the impulse response of the main body filter, a high-quality musical tone can be synthesized without requiring an expensive main body filter. In addition ,
Wherein the excitation signal supplied to the input of the closed-loop means is a signal having a component corresponding to a first partial response in the resonant member or system to be modeled according to the invention, The resonance characteristic according to the second partial response in the resonance system is added to the output signal of the closed loop means by the resonance filter means or the resonance adding means. As described above, since the overall response in the resonance member or the resonance system to be modeled is separated into partial responses and shared before and after the closed loop, the configuration of the excitation unit and the resonance filter unit or the resonance addition unit is required. It can be simplified and its design and manufacturing costs can be reduced.

In one embodiment of the present invention, a complicated and expensive body filter conventionally required can be eliminated, and the size of a table required to store a collective excitation signal can be reduced. it can. By decomposing the resonator into a dump mode and a ringy (ringy) mode, and by setting the collective excitation signal using only the impulse response of the dump mode,
The size of the excitation table can be reduced, and the size of the memory required to store the collective excitation signal can be reduced.

The excitation signal generator may be embodied as a single fixed excitation signal or as a plurality of excitation signals combined to form a combined excitation signal. By controllably weighting each of the plurality of excitation signals, a number of different combined excitation signals can be provided. Furthermore, the various excitation signals can be controlled to change over time, so that significant controllability and musical tone changes can be achieved despite the use of a fixed set of excitation signals.

[0017]

An embodiment of the present invention will be described below with reference to the accompanying drawings. The present invention described below may be implemented in any form of hardware including various delay circuits and filters, or software, for example, using an appropriate algorithm executed in a DSP. It may be.

FIG. 1 shows a conventional delay unit with a filter including a delay element 10 and a filter 12, and a digital main body filter 18 simulating a resonator of a natural musical instrument such as a guitar, that is, a resonance unit. It is a figure showing a delay loop with a filter. An excitation source 14 supplies an excitation signal to the delay loop. The inventor of the present invention has proposed that the delay loop with the filter and the main body filter 18 are
Basically, I realized that it was a linear time-varying system. For this reason, as shown in FIG. 2, without changing the characteristics of the resulting tone,
Delay loop with filter and body filter 18
Can be reversed.

That is, in FIG. 2, the main body filter 18 is provided before the delay loop with the filter. Since the overall processing requirements are the same, this order change itself does not provide a significant advantage.
However, if the variable for simulating the string is selected to be a transverse acceleration wave, as shown in the prior art, the ideal string plucking sound will be an impulse. In this case, the output of the excitation table for plucking the strings is a single non-zero sample, i.e., an impulse, sandwiched between zeros before and after each pluck.
As a result, what excites the delay loop with the filter is the impulse response of the body filter 18.
Since the body filter 18 does not change during the generation of one sound, the body impulse response is fixed. In view of this fact, the present invention intends to completely eliminate the need for providing a main body filter. Instead of passing the impulse through the body filter, the excitation table is loaded with a collective excitation signal representing the impulse response of the desired body filter. In this way, expensive body filters (or filtering in a DSP system) needed to simulate the connection of strings to the resonating instrument body or other coupling structures
Can be eliminated.

FIG. 3 is a diagram showing an example of the configuration of a tone synthesis system according to the present invention. This music synthesis system
In the example shown in FIG. 3, an excitation source including a table 20 for supplying a collective excitation signal e (n) in response to a trigger signal (for example, a key-on signal) 22 is provided. The collective excitation signal is supplied to a delay loop with a filter via an adder 24. The delay loop includes a delay line 26 having a variable length N and a loop filter 28. The output of the delay line 26 is taken as a tone synthesis output x (n) and returned to the loop filter 28 (a number of outputs can be taken as is known in the art). The output y (n) of the loop filter 28 is fed back to the adder 24. The length N of the delay line 26 realizes coarse pitch control.

The loop filter 28 realizes fine pitch control and determines a change of one performance sound. This filter is usually fixed for the duration of one note, but during the generation of one note, it is damped by the player's hand, and the two-stage amplitude envelope attenuation (eg, piano sound) ), Various effects such as amplitude envelope beats due to coupling with other strings, pseudo-reverberation in which a small damped amplitude envelope persists after the nominal cutoff time of the note, and other time-varying effects. Can be changed to occur. The excitation signal determines the initial spectral content of the musical tone, including physical excitation and details due to the body filter where the pick is located when simulating a guitar pluck.

The impulse response (IR) of the loop filter, expressed as f (n), where n is a variable equal to 0, 1, 2,... And the loss due to the coupling of the string to the instrument body. The determination of a particular loop filter characteristic is known and will not be described in detail here. The impulse response f (n) is obtained by an equation relating to basic physics concerning theoretical loss due to connection or coupling between the string and the body. The material of the string,
Using the tension and the diameter, the loss per unit length of the string can be estimated theoretically. For example, the loss at a body connection point such as a guitar bridge can be estimated from the shape of the bridge and the resonance of the instrument body. Further, the impulse response f (n) can be obtained from a physical measurement value of a string of an actual stringed instrument. Further, a combination of mathematical expressions and estimates based on actual physical measurements may be used.

The excitation signal e (n) can be set using a number of different methods. This excitation signal e
(N) is set both by the nature of the physical excitation of the string and by the response of the instrument to the point of excitation by the string. For example, in the case of a guitar, the excitation for the instrument body is
Occurs at the guitar bridge. FIG. 4 is a physical block diagram of the guitar. In this figure, when an excitation signal 30 is applied to a string 32, the string 32 excites a resonance unit (guitar body) 34. In a physical system, the resonating part is set by picking out the output signal. In a typical example, the output signal is selected at a position several feet away from the front plate of the guitar body. In fact, such a signal can be measured by using a microphone held at the desired output point and recording the response at that output point to tapping the guitar bridge with a force hammer. In addition,
The resonance section includes not only the resonance characteristics of the guitar body itself,
It also includes air transmission characteristics. If an output point far from the guitar is selected in the reverberation room, the resonance characteristics of the room in which the measurement is made are also included. FIG. 5 shows the collective characteristics of such a resonator.

The overall resonance section 34 is a bridge connection (br
idge coupling) 36, guitar body 38, air absorption 40
And a room response 42. Generally, it is preferable to select an output relatively close to the guitar so that the impulse response of the resonance section can be as short as possible. However, the universality provided by the ability to incorporate all downstream filtering into a single resonator is an important feature of the present invention. This is more evident in the case of the piano model of FIG. 6, where the resonating plate and the enclosure are combined as one resonating part. In this case, the overall resonating part 34 includes a bridge connection 44, a piano resonating plate 46,
Consists of a piano enclosure 48 and an air / room response 50.

The only technical requirement for the components of the resonator is that these components have a linear and time-varying characteristic. As mentioned above,
These two properties mean that these elements may be provided in any arrangement order. If the string also has a linear characteristic and a time-varying characteristic, the arrangement order of the resonating part and the string may be reversed as shown in FIG. In fact, said strings are the least linear element of almost all stringed instruments, but the main effect of their non-linearity is that the fundamental vibration frequency increases slightly with amplitude. For the purpose of reversing the arrangement order, the strings can be considered to have sufficient linear characteristics. The strings have a time-varying characteristic even in the presence of vibrato, which is also a secondary effect. The consequences of changing the arrangement order of the slowly time-varying strings and resonating parts are not mathematically the same, but the musical tones generated are basically the same.

After reversing the arrangement order of the strings and the resonance section as shown in FIG. 7, the next step is to combine the excitation signal and the resonance section as shown in FIG.
A collective excitation signal 52 is obtained. This collective excitation signal 52 is set to provide an output a (n) that is basically the same as the output of the resonator shown in FIG. For this purpose, first, the excitation characteristics must be specified. The simplest example is an impulse response. Physically, this would be the most appropriate choice if strings were used to model the acceleration wave. In the case of this example, an ideal plucking generates an acceleration impulse input to the string. In this simple example, the collective excitation signal 52 is simply the sampled impulse response of the selected resonator.

In a more complex example, the excitation signal is e
(N), when the impulse response of the resonance unit is r (n), the equivalent collective excitation signal a (n) is given by the following equation (1).
Given by convolution of e (n) and r (n).

(Equation 1)

When the collective excitation signal is long, it is desirable to shorten it using some technique. To this end, it is useful to first convert the signal a (n) to a minimum phase, as described in various references relating to signal processing. In this way, the maximum shortening that matches the original magnitude spectrum is realized. The signal a (n) can then be windowed, for example, by using any suitable part of the various window functions used for spectral analysis. One example of a useful window is an exponential window. The reason is that the exponential function window has an effect that the damping rate of the resonance part can be uniformly increased.

As shown in FIG. 9, the excitation signal is obtained by recording a sound generated from the musical instrument (for example, a sound of striking a string) and performing an inverse filtering process to remove a component due to a string loop. May be set. In FIG. 9, the string loop filter is set by one of various methods,
Included in the inverse filter. The resulting output includes components corresponding to the pluck and body filters and can be used as an excitation signal (or as a reference signal to obtain a modified excitation signal).

FIG. 10 is a diagram showing an impulse response of a typical main body filter of a natural musical instrument. Basically, this impulse response is a dump oscillation waveform. It is such a response that is stored as a collective excitation signal in the simplest case where the excitation signal is an impulse. In other instances where the excitation signal is other than an impulse, the aggregated excitation signal will be the convolution result as described above. Since this convolution is due to the impulse response, the convolution result ends with a dump oscillation waveform in any case. However, various shortening techniques provide an excitation signal having a waveform other than the dump oscillation waveform. Such a shortened excitation signal is obtained from the original impulse response (and provides a similar result to the original impulse response).

The tone synthesis system can be used to simulate the input of an excitation signal at different pick locations, ie, different locations along a string. By simultaneously exciting the string at two different positions along the delay line and adding to the component present at that point in the delay loop, a particular pick position on the string is simulated. This is shown in FIG. 13, where the delay circuit is divided into two delay circuits 54, 56, between which adder 58
Is inserted. In general, the ratio of the delay time at the pick position to the delay time of the entire loop is equal to the ratio of the pick position to the length of the chord. Delay circuits 54 and 56
The total delay length N corresponding to the selected pitch corresponds to a desired tone period corresponding to the selected pitch (however, N = “delay amount corresponding to pitch” − “delay amount of loop filter”). Here, the delay amount P of the delay circuit 54 can be changed corresponding to a desired pick position, and accordingly, the remaining delay amount NP of the delay circuit 56 is changed.

FIG. 14 shows an example of a technique related to the above, in which the excitation signal is delayed and added to an undelayed excitation signal to achieve basically the same effect as in FIG. In FIG. 14, unlike FIG. 13, a pick position delay circuit 60 and an adder 62 are provided separately from the delay loop. As before, the pick position delay circuit 60 can be varied to control the actual pluck point on the string.

The tone synthesis system according to the present invention may be modified to provide a large number of excitation signals in order to realize the effects of many sound emission points in a natural musical instrument. A listener who listens to wooden and metal instruments receives signals from many sound emitting surfaces on the instrument. Therefore, different signals reach both ears. Further, as the player moves the instrument or the listener moves his head, the mixed sound emitted from the instrument changes dynamically. In order to handle such natural phenomena, it is useful to be able to generate many output signals corresponding to different output signals in a natural environment. According to the invention, this is easily simulated by providing a number of collective excitation signals, each having a different component reflecting a different body filter or a different overall resonant system, as shown in FIG. It is possible. In FIG. 15, collective excitation signals 64, 66 are provided and provided to a single string delay loop 68 (if separate outputs are desired, separate string loops may be provided). Although only two collective excitation signals are shown, any number of excitation signals may be provided to simulate crossfading at multiple different output points. Interpolation between two or more tables may be used. Two or three
One or more set excitation signals 64 and 66 may be interpolated as appropriate and applied to a single string delay loop 68.

An important change in the tone synthesis system is that the excitation table is read semi-periodically.
Instead of providing a single trigger signal to initiate string plucking, the trigger signal is provided periodically (or approximately periodically to account for vibrato). In this example, the amplitude of the excitation signal is adjusted to provide an appropriate output level (eg, by shifting the output value of the table to the right, or by adding an amplitude envelope to the output value of the table). Can be smaller. This technique can simulate very high quality bowed strings.

If a trigger signal is generated while the excitation table is being read, two variants are possible. First, the excitation table may be restarted from the beginning, thus interrupting the ongoing reproduction. This is shown in FIG. In addition, the reproduction start of the new excitation table starts as shown in FIG.
It may overlap with the ongoing playback. This variant is more complicated since it requires a separate execution pointer and adder for each regeneration of the excitation table.
However, this is more preferable in terms of quality.

In a useful variation, in addition to providing a mixed excitation signal as in FIG. 15, a plurality of excitation signals (tables or other) are provided as shown in FIG. Gain control is provided for each variable excitation signal. In FIG. 16, the excitation signal generator 70
Generate the excitation signals. Each excitation output has a gain control element 72 that can be changed over time. The outputs of these gain control elements 72 are combined by an adder 74 to provide a collective excitation signal a (n).
This signal is provided via an adder 80 to a delay loop including a delay line 76 and a loop filter 78.
Providing a gain control element for each excitation signal in this manner provides a means for combining a wide range of excitation signals as a time-varying linear combination of fixed excitation signals. That is, each excitation signal is fixed, but by controlling the relative gain of each excitation signal, the relative contribution to the total excitation signal provided to the delay loop can be controlled. The gain may be set to a specific value and held for the duration of one note, or to further change the characteristics of the tone being generated in addition to the change due to the filter-added delay loop itself. , May be changed over time.

Typically, the gain gi (n) is fixed so that in a free vibration, such as a plucking sound, only one linear combination of the excitation signals is used. On the other hand, for example, in the case of a driving vibration such as a bowed string, the gain gi (n) can be changed over time in order to change the characteristics of the musical sound. This can be achieved by providing a smoothly varying envelope for each excitation signal and controlling the relative contribution of a plurality of different excitation signals. The time change realized by changing the excitation signal is in addition to the time change realized in the delay loop with the filter.

The characteristics of the various excitation tables can be selected to maximize the number of useful changes that can be achieved from a fixed set of tables. For example, a set of excitation tables may include a number of waveform tables stored in ROM in addition to a noise generator with a filter. The waveform table can provide various aggregated excitation signals taking into account different body filters, or the main components (eg, frequencies) of the overall desired excitation signal are provided separately in different waveform tables. , May be based on an analysis of the main elements that are variably combined. This is similar to the well-known Fourier synthesis used for standard tone generation (but not for the excitation signal generation for delay loop tone synthesis).

The tone synthesis system shown in FIG. 16 is useful for simulating the sound of a bowed string.
In general, accurate simulation of such sound requires a delay loop that has a non-linear junction to capture the excitation signal and the signal circulating through the loop, and feeds back the signal according to a non-linear function. is there. However, the tone synthesis system of FIG. 16 does not require the non-linear junction, and nevertheless uses only a filtered delay loop and a time-varying excitation signal to create a bowed string. High quality simulation can be realized. In this regard, each excitation signal itself is time-varying but has a relatively short fixed duration. To generate a sustained tone, such as a simulated bowed string, each excitation signal is repeated multiple times, and a temporal change in the relative intensity of each excitation signal results in the desired tone change.

FIG. 17 shows a modification which provides a significant computational advantage. Generally, the initial attack portion of a musical tone contains significant high frequency information. In order to properly synthesize the attack portion in a normal delay loop with a filter, the sampling rate of the loop filter must be maintained at a relatively high rate. This does not apply to other parts of the synthesized tone that have less high frequency components.

As shown in FIG. 17, the present invention reduces computational requirements by providing a separate attack signal as one of the excitation signals and bypassing it around a delay loop with a filter. is there. The attack signal includes a high-frequency signal having a short duration (for example, 100 ms) that is read in parallel with another excitation signal in response to a trigger signal. In FIG. 17, the attack signal is denoted by reference numeral 8.
It is provided at two locations, is gain controlled by an amplifier 84, and is provided to an output summing junction 86.
An additional excitation table 88 is appropriately weighted at 90 and summed at 92 to provide a combined excitation signal e (n). The combined excitation signal e (n) is supplied to the delay line 94 and the loop filter 96.
And a delay loop with a filter including the adder 98.

Since there is no need to process high frequency components, the sampling rate in the loop filter 96 may be very low. For example, in the case of generating a low pitch sound such as a low E sound of a guitar at low cost, the excitation signal input to the string loop may be limited to 1.5 kHz,
First 100 ms of recorded sound high-passed at 1.5 kHz
May be used for the attack signal. Also, a sampling rate of 3 kHz may be used for the delay loop.
The output signal of the loop is 22 kHz by the interpolation circuit 100.
may be upsampled to z and added to the attack signal (also provided at a sampling rate of 22 kHz). Although the composite excitation signal z (n) contains both the desired high and low frequency components, nevertheless the processing of the delay loop is greatly simplified. The sampling rate of the string loop may be controlled as a function of pitch.

The synthesizing technique of the present invention is also applicable to vibraphones having a small number of exponentially damped resonance modes, and to musical sound synthesis of other percussion instruments such as tom-tom, marimba, and iron harp. In these cases, the outputs of the plurality of delay loops with filters are summed up, whereby the most important resonance mode can be imitated as the sum of a series of substantially harmonic oscillation modes. This technique is also applicable to wind instruments. The excitation table in this case supplies an impulse response from the inside of the wind instrument tube to the outside of the sound hole and the morning glory portion. Natural articulation is difficult to obtain because there are no non-linear junctions (typically used for physical simulation of wind instruments) that provide an interaction between the musical waveform and the excitation signal . However, this technique is simple and inexpensive to implement.

FIGS. 18 to 30 show another embodiment of the present invention. In this embodiment, the size of the excitation table can be reduced and the cost can be reduced by dividing the resonance section into the "dump" mode and the "ringy" mode. In this case, the resonance part with the least dump is extracted, and only the remaining resonance part with more dump is replaced with the string in the arrangement order.

As described above with reference to FIGS. 7 and 8,
In the simplest example, the collective excitation table 52 is essentially the sampled impulse response of the selected resonator (eg, the guitar body). In a more complex example, the collective excitation signal is provided by performing the convolution shown in equation (1) above, where the impulse response of the resonator is convolved with the excitation signal e (n).

The length of the convolution result, and thus the length of the impulse response r (n) of the resonator, which affects the collective excitation signal stored in the table, is determined by its least damped resonance. I am trying to. The inventor has determined that from the more dumped resonances to the least dumped resonances, ie, the longer ringing (or ringing) mode (ie, Ringy mode).
It has been found that by eliminating the above, the resonating portion, which is swapped in arrangement order with the strings, will only have a portion that is more dumped. This resonating part has a shorter impulse response.

The long ring portions, whose placement order is not interchangeable, can be simulated by a small number of two-pole filter sections or other recursive filter structures. It should be noted that the present invention is by no means limited to implementation with digital filters, and may use any suitable digital or analog filter. Since most synthesizers today use a number of "additional" filters to provide a post-processing effect on the generated tone signal, the present invention provides a "ringy" portion of the resonator. By using these filters that work, the excitation table required to store the collective excitation signal can be greatly simplified.

FIG. 18A is a block diagram showing a tone generation mechanism of, for example, a guitar. In this example, when a trigger signal is supplied to the excitation source 30, the excitation source 30 becomes a string. An excitation signal e for exciting the section 32
(N). The string portion 32 outputs a final output signal x
An output signal s that excites the resonance unit 34 that generates (n)
(N). The characteristics of the resonance section 34 are the same as those described above with reference to FIGS.

Instead of immediately replacing the resonating part 34 and the string 32 as in the above-described embodiment, the characteristics of the resonating part 34 are, as shown in FIG.
It is divided into a “dump” resonance section 102 and a “ringy” resonance section 104. Typically, the resonance unit 34 first
It is studied in the form of a measured impulse response. The measured impulse response is, for example, a force hammer,
It may be obtained by using two-channel A / D conversion and system identification software available in the MatLab ™ programming environment known to those skilled in the art.

One of the two A / D conversion channels is for recording a force hammer output proportional to the hitting force of the force hammer. Also, the other A / D
The conversion channel records, for example, a microphone output that measures a response of the resonance unit to operation by the hammer. The system identification software basically estimates the measured impulse response by deconvolving the force hammer "input" signal from the measured microphone "output" signal.

A simple technique for implementing the deconvolution function is to divide the "output" Fourier transform by the "input" Fourier transform, thereby obtaining the measured frequency response of the resonant section. It is to get. Alternatively, a more sophisticated deconvolving process may be performed using a commercially available software package. The impulse response is obtained by performing an inverse Fourier transform of the frequency response using any of the above techniques.

After the impulse response of the resonance section 34 is determined, the most ringing mode of the impulse response is converted to a "parametric form". That is, the exact resonance frequency and resonance bandwidth corresponding to each of the narrowest "peaks" in the frequency response of the resonator is identified and transferred to the "ringy portion" 104. The longest ringy mode corresponds to the narrowest bandwidth. Also, this longest ringy mode typically includes the highest peak in the frequency response.

Thus, an effective technique for measuring the resonance with the longest ring time is to determine the exact location and bandwidth of the narrowest and highest spectral peaks in the measured frequency response of the resonating section 34. That is. The center frequency and bandwidth of the narrow frequency response peak determine the two poles in the ringy portion 104 of the resonator. Representing a filter in terms of its poles and zeros is a type of "parametric" filter notation, as opposed to a "non-parametric" representation such as an impulse response or frequency response.

As is known to those skilled in the art, commercially available system identification software products are available that include software for converting measured frequency response peaks into parametric form. Such commercially available system identification software products include a force hammer and complete data collection means. Moreover, those skilled in the art are familiar with the signal processing literature that has written on this subject. As an example, "Prony'smethod" is a classic technique for estimating the frequency and bandwidth for the sum of exponentially decaying sinusoids (impulse response of a two-pole resonator). A more advanced recent technology is the “matrix pencil
method (matrix pen method) ".

FIG. 19 shows a ringy portion 10 of the resonance section 34, which was executed using a small-scale Matlab program.
4 illustrates a method of converting 4 to parametric form. For simplicity of illustration, only one frequency response peak is shown in this example. First, the center frequency of the peak is measured using a peak finder for quadratic interpolation that operates based on the spectrum scale in decibels. Next, in order to design a two-pole filter having a frequency response that mimics the measured data as closely as possible, a general-purpose filter design function “invfre
gz () ”is called.

To determine the parametric filter coefficients (as performed in the example of FIG. 19), a known "equation-error method" for designing a digital filter is used.
(Equation error method) ", so that the filter design program focuses on the spectral peaks, as also shown in FIG. 19 (after renormalization to be overlaid on the figure). A weighting function is used, the weighting function used in this example is 0H
It is “1” in the range from z to 900 Hz, “100” in the range from 900 Hz to 1100 Hz, and then returns to “1”. The weighting function in FIG. 19 appears as a rectangular function centered on the 1000 Hz spectral peak.

FIG. 19 shows a size-frequency-response overlay of a two-pole filter designed by the above-mentioned equation error method. As shown, the fit between the non-parametric and parametric frequency response is very high near the peak. By using the initially measured interpolated peak frequency, the polar angle of the desired filter can be fine-tuned, thus making the equation error technique a technique for measuring only the peak bandwidth in this case. Can be. As known to those skilled in the art, there are numerous techniques for measuring spectral peaks in the field of signal processing, and the invention is not limited to use with the techniques shown. .

Another method of converting the ringy portion 104 of the resonator 34 to a parametric form is to use a well-known linear predictive coding (LPC) technique followed by a polynomial factorization to determine the poles of the resonator. Is what you do. The LPC is particularly excellent for simulating a spectral peak. The pole closest to the z-plane unit circle is selectable for the ringy portion 104 of the resonator 34.

If the LPC or any other "minimum phase" parametric form is used to implement the ringy portion 104, the corresponding "dump"
Portion 102 uses a process called "inverse filtering", which is a process well known in particular for linear predictive coding and system identification, to provide the full impulse response of the resonator 34 and the parametric or ringgit It can be calculated from part 104.

The inverse filter is constructed by providing an all-zero filter whose zeros are equal to the poles of the ringy portion 104. If the ringy portion 104 has zeros, these zeros must be stable because they become the poles of the inverse filter.
In the case of a digital filter, the zero must have a magnitude less than 1 in the z-plane. In the case of an analog filter, the zero point must be located on the left half of the s plane. Such filters are known as “minimum phase”
It is called a filter. To reduce the likelihood of obtaining a non-minimum phase in the estimated parametric form of the ringy portion 104, the initial impulse of the resonator may be reduced using a non-parametric scheme such as the known keptoslam "convolution" technique. It is useful to convert the response to a minimum phase impulse response.

In both digital and analog filters, if the zeros are non-minimum in phase, these zeros will be around the appropriate frequency axis so that a minimum phase filter with the same magnitude of frequency response can be realized. Must be reflected in Thus, the inverse filter is applied to the full impulse response of the resonator to obtain a "residual" signal. This residual signal is an impulse response of the “dump portion” 102 and is suitable for rearranging the arrangement order with the strings and convolving with a string excitation signal such as a “plucking sound” signal. When the residual signal is supplied to the ringy portion 104, that is, the parametric resonator (in this case, the minimum phase filter), the original impulse response of the resonator 34 can be realized with high accuracy. In general, the accuracy in this case is simply affected by numerical roundoff errors that occur during the backward and forward filter operations.

The present inventor has determined that the all-pole filter is convenient and easy to operate. All-pole filters are always minimum phase, and the LPC technique easily calculates them. As will be appreciated by those skilled in the art, there are many filter design techniques capable of generating a parametric portion having a predetermined number of poles and zeros, and using a weighting function to ring the method to the longest of the impulse response 34. Can lead to ingredients. The equation of error shown in FIG. 19 is an example of a method that can calculate not only poles but also parametric, ie, zeros in the ringy portion. Thus, the parametric portion 104 may have any number of poles and zeros and may be implemented using any known filter implementation techniques.

[0063] Known digital filter implementation techniques involve serial and parallel connections of the secondary filter sections. As is known in the art, the transfer function of any linear time-varying (LTI) filter can be incorporated into a series connection of rudimentary secondary parts. Also, as is also known in the art, all LTI filters are divided into parallel sums of quadratic parts by a "partial fractional extension" operation. Each said secondary part is capable of resonating at only one frequency or not resonating at all.

FIG. 20 shows an embodiment of a general second-order filter section called “Direct Form I” (up to two poles, up to two zeros and one gain factor). I have. Although there are several other embodiments, this Direct
Form I is a numerically preferred choice because the output of all multipliers is sent to a common adder that typically uses two's complement arithmetic. As a result, the overflow only occurs in one digit of the output. To obtain the highest quality, the feedback signals (all of which in the figure are to the right of the scaling factors b0 to b2) can be implemented with twice the precision. If the excitation table is scaled accordingly, the coefficient b0 in FIG. 20 can be removed. In the figure,

(Equation 2) Realizes a delay of one unit (one sampling cycle),
It is a unit delay element.

When the secondary filter resonates, its resonance frequency and resonance bandwidth are determined by the feedback coefficients a1 and a2 according to the following equations. a1 = -2 R cos (2 Pi Fr / Fs) (2) a2 = R 2 (3) where, Fr is a value representing the resonant frequency in cycles / sec That hertz (Hz), Fs is a digital audio The sampling rate is expressed in Hz, Pi is 3.141 ...
Where R is the pole radius associated with the resonance bandwidth by the following equation: R = exp (-Pi Br / Fs) (4)

The time constant of the damping Tr for the secondary resonance section is related to the bandwidth by Tr = 1 / (PiBr) (5).

The time constant is defined as a time (second) in which the impulse response of the resonance section attenuates at a rate expressed by 1 / e = exp (-1).
Is defined as In this embodiment, it is necessary to identify the "most ringing" secondary, that is, the part with the longest decay time Tr (or the smallest bandwidth Br). These parts may obviously be implemented as secondary parts of the parametric part 104 of the body resonance part 34.

FIG. 21 shows a less common secondary resonator having only two poles. The parametric portion 104 is selected as an all-pole type,
This configuration can be used when the secondary parts are connected in series. FIG. 22 is a diagram showing a series connection of two two-pole parts. Perhaps this is the most convenient choice, but the inventor has shown that this is a two-part parallel connection in FIG.
Not as good in numerical performance as the 3 example,
I discovered that. As will be appreciated by those skilled in the art, a suitable fractional extension of the transfer function will result in parallel quadratic parts each having at most one zero and up to two poles.

In general, when a digital filter produces disjoint resonances (ie, resonances do not overlap in the frequency domain), it is numerically preferable to use a parallel secondary rather than a series secondary. preferable. This takes into account that if a series secondary is used, it is also necessary to compensate for the signal attenuation by all other non-resonant filter sections in order to obtain a resonance peak at a certain frequency. You can understand by doing. on the other hand,
When a parallel secondary part is used, the resonating secondary part basically performs resonance operation independently. Thus, in general, the parallel secondary shown in FIG. 23 is numerically superior to the series secondary shown in FIG. However, a parallel secondary is less convenient to calculate than a serial secondary, and requires a zero for each secondary to properly phase the outputs.

The effects processor provided in most commercially available tone synthesizers usually includes a "parametric equalizer unit". Typically, each of the parametric equalizer units is limited so that b0, b1, and b2 realize one gain control.
Is a secondary resonance section as shown in FIG. Usually, the parameters of the equalizer section are a center frequency, a bandwidth and a gain for each equalizer section. For this reason, the parametric equalizer section usually used for adjusting the mixing of various frequency bands in the synthesized musical tone can also be used to realize a desired ringing mode of the main body resonance section.

When the disassembly of the resonance section is completed and the parametric section, that is, the ringy mode section 104 is realized,
As shown in FIG. 18C in the above-described embodiment, the most dumped resonance of the resonance section 34 is replaced with the string 32. Next, in order to generate the collective excitation signal 106 as shown in FIG. 18D, the dump resonance unit 102 is convolved with the excitation signal 30 using, for example, the above equation (1).

As described above, in this embodiment,
A trigger signal is provided to the collective excitation signal 106, which excites the string 32 via the excitation signal a (n). Thus, the string 32 processes the input signal,
An output signal r (n) that does not include the long ring component of the resonance section 34 is generated. On the other hand, these long-ring components may be configured on the output side of the string 32 by resonant filters, for example a number of two-pole filter units, connected in series or in parallel depending on the nature of the signal being synthesized. It is supplied via the ringy mode resonance unit 104. As a result, the signal output from the resonance unit 104 becomes a tone signal.

An additional advantage of having a separate parametric or ringy mode resonator 104 is that only one output signal is readily available in the unresolved resonator 34, whereas Is available. These plurality of output signals can be used to variously improve the quality of a synthesized musical tone. As an example, the outputs of the secondary resonators connected in parallel can be "panned" stereoscopically at various locations. This panning can be selected to mimic the spatial distribution of the resonant modes of the simulated instrument. By slightly altering this stereo arrangement, a spatially moving instrument can be simulated.

In order to realize various stereo arrangements of the individual resonating portions of the parametric resonating portion, ie, the ringy mode resonating portion 104, one adder for taking in two resonating portion outputs shown in FIG. It is replaced by two adders, one for the "left channel" and the other for the "right channel". Further, for each of the resonance units,
Two scaling means are used, for example multipliers, one scaling means scales the output signal before being summed to the left channel and the other scaling means scales the output signal before summing to the right channel. . By adjusting the respective scaling factors, the amount of the output signal sent to the left and right channels will determine the stereo placement of the signal.

When two scaling means are used as described above, the "b" coefficient of each resonance section (for example, FIG. 23)
B0a and b0b) can be omitted. Thus, only one additional multiplier is required for a stereo arrangement. Also, since the angle of the stereo configuration is often not a critical factor, the two scaling factors are 0, 1/8, 1/4, 3
It may be a specially quantized number that does not require multiplication, such as a number that can only assume the values / 8, 1/2, 5/8, 3/4, 7/8, 1. For example, multiplication by these numbers can be performed (with binary fixed point) using one or two shifts and zeros or one fixed point addition or subtraction.

The size reduction of the excitation table which can be realized by using the technique according to the present embodiment is achieved by observing an ideal composite impulse response of the guitar end before and after one mode having the smallest dump amount is excluded. Can be illustrated by doing so. FIG. 24 shows the initial impulse response of a simulated guitar body resonating at 100 Hz. As is known, in guitars, the main resonance of the guitar body generally provides the longest ringing resonance, and therefore, a ringing element with the least amount of dumping of the body impulse response as shown in FIG. appear. As can be seen from FIG. 26, the excitation table can be shortened by one digit by removing this single, secondary, and least-resonant 100 Hz resonance component. In this example, the component to be removed can be created, for example, by a single secondary resonance filter 104 shown in FIG. 21 having a resonance frequency of 100 Hz.

FIGS. 27 to 29 are views showing similar examples using data measured from an actual classical guitar. FIG. 27 shows the estimated impulse response of the resonator 34 converted to a minimum phase using the known Keptoslam "convolution" technique. In this case, one is 110 Hz and the other is 220
There are two long ringing low frequency resonances in Hz.
Since there are two ring resonances, each producing a spectral peak in the frequency response, the parametric resonator 104 needs to have at least four poles.
FIG. 28 shows the impulse response of the four-pole parametric resonator 104 calculated using the equation error method. Although the backward filtering has been performed, FIG. 29 shows the residual impulse response 102. The small noise bursts that appear at intervals of about 12 msec are related to the pitch of the guitar strings that were extruded to make the measurements shown and are not relevant in this example.

The reduction in the size of the excitation table which can be realized according to this embodiment considerably saves the overall cost of the tone synthesizer. Further, this embodiment implements a "ringy" mode resonator using a relatively simple resonant filter, and has the additional feature that the filter is already present in most synthesizers currently in production. No cost is required.

The cost savings realized by extracting the longest ringing resonance mode of the impulse response of the resonance unit 34 to the parametric unit 104 are, among other things,
(1) the duration of the impulse response; (2) the duration of the impulse response remaining after extracting the longest ringing resonance mode; (3) the cost of the memory;
(4) It depends on the cost of realizing the secondary filter unit. According to current hardware trends, processors are provided that perform faster processing in increasingly smaller configurations where only a small amount of memory is available locally. Reducing memory usage at the expense of more processor utilization is a welcome trade-off.

In this manner, this embodiment can provide an effect equivalent to the overall effect of the first embodiment, that is, eliminate the need for an expensive and complicated main body filter which has been conventionally required. However, in this embodiment, by extracting the resonating part into a component based on its dump component, it is possible to maintain a resonating part that can be easily simulated using a resonant filter, and a much more complex dump The section can be convolved with an excitation that produces a collective excitation signal that provides a downstream resonance characteristic.

Further, this technique removes most of the large-sized resonance part and reduces the size of the excitation table while utilizing the capability of the synthesizer by realizing the ringy mode resonance part using the existing resonance filter. By reducing the size, the tone synthesis processing can be simplified. This technique can be used with other embodiments including the use of multiple excitation tables as shown in FIGS.

Although the embodiment described above with reference to FIGS. 18-29 can also be used with the delay loop of the above embodiment shown in FIGS. 3, 12, etc., FIG. An improved filtered delay loop capable of simulating a detuned string is shown. As shown in FIG. 30, an input signal (for example, a
(N)) includes an adder 108 and a one-cycle delay element 110
Given to.

The delay element 110 produces two outputs. The first output is a mobile interpolated tap 112, ie, an output that is delayed from the continuously changing position along the line of the delay element 110 by an amount proportional to the removal point. . This output is a scaling factor g, which may be a time-varying value.
Can be scaled by The output of the delay element 110 is also provided to an adder 114, which feeds back to a low pass filter 116 and subsequent adder 108. The output of the movable tap to be interpolated 112 is
Downstream of the delay loop and outside the delay loop.

With the above configuration, several features of the string synthesizer can be realized which have the effect of requiring high cost. First, a flanged chord can be realized at low speed using a back and forth movable interpolated tap. Ideally, a large number of independently moving taps will provide the best flange effect (eg, as shown by the dashed lines in FIG. 30). Each of the taps adds moving comb filtering to the output. A single non-movable tap can provide the fixed comb filtering required to simulate the position of a pluck, tap, or other excitation operation along the string. In this case, the non-moving taps do not require interpolation since the exact location of the physical excitation is not fully audible.

In addition to simulating flanged strings, a detuned "second string" can be simulated by using faster one-way taps. In this case, the speed of the tap corresponds to the Doppler shift. The faster the moving tap moves to the right in FIG. 30, the slower the Doppler shift of all frequencies in the input signal. On the other hand, the faster the movable tap moves to the left in FIG.
The Doppler shift of the frequency is faster. In this embodiment, when the mobile tap reaches the end of the delay line, the tap must somehow "circulate" to the other side.

In the simplest case, a simple circulation method can be used. By cross-fading so that the output of one tap that has reached the entrance end of the delay line fades in and the output of the other tap that has reached the exit end of the delay line fades to zero at the same time, A better sound is obtained. Thus, in this case, during the crossfade, two mobile, non-interpolated taps are active. A more elaborate cycle is to look for a suitable jump along the delay line. For example, if possible, the wraparound may be performed at a zero crossing point, or the correlation at various delay points may be calculated. All of these technologies are
It is a technique that has been known to some extent in the context of "harmonizer" and "pitch shift" algorithms. By adding taps of different tap speeds, other detuned strings can be simulated. By creating a large number of virtual strings in this way with slightly different tunings that change over time,
A pleasant "chorus effect" is obtained.

Flanging and Doppler shift can be used to simulate the vibrating string coupler effect. Such coupling produces a slow beat in the amplitude envelope of the overtones of the ringing sound.
Mobile comb filters (with notches that are not too deep) can have a qualitatively similar effect due to flanging. Alternatively, an equivalent effect is achieved by adding the output of the string to a virtually detuned string. As a specific example, if the scaling factor g in FIG. 30 is set to 0.25 and the tap speed is set to produce a Doppler frequency shift of 0.25%, the sum between the two slightly mistuned strings will be Produces a beat similar to that found on electric guitars.

The simulation of the multi-string coupler sound required in almost all stringed instrument sound synthesis is as follows:
Using the present invention is cost effective. Conventionally, a plurality of string simulators have been required to simulate a similar multi-string coupler sound.
However, in the present invention, using only one string simulator, the multi-string coupler effect is added by one or more mobile taps.

As a modification of the above-described string coupler simulation, a string coupler with two filter-connected delay loops each simulating the string shown in FIG. 31 may be used. In this variation, the outputs of each filtered delay loop are summed and the summed signal is scaled using a negative coefficient, preferably having a magnitude of less than or equal to about 0.01. As shown in FIG. 31, in order to perform a more accurate simulation, the negative coefficient is replaced with a transfer function −Hb (z) that can be calculated from a measured or theoretically predicted connection characteristic by a filter. Can be The signal thus scaled or filtered is feedback-added to each said filtered delay loop via a feedback path, preferably
The output is introduced into the loop at a location immediately after the location taken from the loop. The output used for coupler sound may be taken from any location in the loop.

Such a genuine coupler method is performed as follows for N strings. (Corresponding to N strings)
The outputs of the N filtered delay loops are summed, and the summed signal is scaled by "-epsilon" (or filtered by -Hb (z)), and the scaled or filtered signal is Preferably, it is introduced into each said loop via a feedback path. The scaled (or filtered) signal represents a physical interpretation of the "bridge" output. The scaled signal, or, in most cases, the unscaled sum signal, provides an excellent choice for the aggregate output of the connected string assembly.

Unlike the case where negative coefficients are used to implement a true coupler, when a filter of -Hb (z) is used, all interconnected (coupled) filters are used. The loop filter in the attached delay loop can be eliminated. That is, the coupling filter -Hb (z) can be used to provide all the filtering required for all connected delay loops with filters. If a separate loop filter is not used, the coupling filter can be considered a shared loop filter.

Further, these effects (ie, flanges, detuned strings, choruses, virtual couplers, true couplers, etc.) can be used for any combining technique using a filtered delay loop. For example,
There is a wave guide synthesis of wind instruments, brass instruments and percussion instruments.

In summary, according to the present invention, by providing an excitation signal corresponding to a triggered (optionally processed) impulse response, the need for expensive and complex body filters is eliminated. High quality “nailed”, “slammed” and “bowed”
Can synthesize musical sounds. The characteristics of a resonant system provided downstream of a vibrating element such as a string are provided by appropriately generating an excitation signal that takes into account the impulse response of the downstream resonant system.

The configuration of the tone synthesis technology according to the present invention is greatly simplified as compared with the conventional system requiring a complicated main body filter. Further, by disassembling the resonance section into a dump mode and a ringy mode, and then replacing the dump mode with a string and convolving with an excitation signal, it is possible to remove a complicated and expensive main body filter conventionally used, and to reduce an excitation table. Can be reduced in size, and costs associated with the excitation table can be reduced. Flange and chorus effects, as well as virtually detuned strings, can be implemented simply by moving along the delay line of the filtered delay loop and adding interpolated taps that are summed with the output.

[0095]

As described above, since the present invention does not require an expensive and complicated main body filter, the present invention has an excellent effect that high-quality musical sounds can be synthesized easily at low cost.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a conventional tone synthesis system based on a delay loop with a filter using a main body filter.

FIG. 2 is a block diagram showing a musical sound synthesis system in which the arrangement order of a delay loop with a filter and a main body filter is changed.

FIG. 3 is a block diagram showing a tone synthesis system according to an embodiment of the present invention, to which a collective excitation signal corresponding to an impulse response of a main body filter is provided.

FIG. 4 is a block diagram showing a tone generation mechanism of a guitar.

FIG. 5 is a block diagram showing generation of a musical sound by the guitar and the surrounding space.

FIG. 6 is a block diagram showing a tone generation mechanism of a piano including the surrounding space.

FIG. 7 is a block diagram showing an equivalent tone generation mechanism in which a resonance section is arranged in front of a string.

FIG. 8 is a block diagram showing a tone generation mechanism in which a response of the resonance unit to a specific excitation is used as a collective excitation signal.

FIG. 9 is a block diagram showing an inverse filter method for setting an excitation signal.

FIG. 10 is a diagram illustrating an example of an excitation signal corresponding to an impulse response of a main body filter.

FIG. 11 is a diagram showing an example of an excitation signal repeatedly supplied to realize sustained musical tone generation.

FIG. 12 is a diagram showing another example of an excitation signal repeatedly supplied to realize sustained musical tone generation.

FIG. 13 is a block diagram showing a musical sound generation system capable of simulating a change in the position of a pick.

FIG. 14 is a block diagram showing an equivalent system for changing the position of a pick.

FIG. 15 is a block diagram illustrating a system using two excitation tables that are scaled and summed to generate a final excitation signal.

FIG. 16 is a block diagram showing a tone synthesis system incorporating a time-varying mixed excitation signal generator.

FIG. 17 is a block diagram showing a tone synthesis system incorporating an excitation signal generator for generating an attack component outside a delay loop.

FIGS. 18A and 18B are block diagrams showing a tone generation mechanism of a guitar, wherein FIG. 18A shows a general example of the tone generation mechanism, and FIG. (C) shows an example in which the dump mode unit is provided in front of a string, and (D) shows an example in which an excitation signal is folded with the dump mode unit.

FIG. 19 shows a non-parametric frequency response, a parametric frequency response fit and an overlay of a weighting function.

FIG. 20 is a diagram showing a configuration example for realizing a digital resonance unit.

FIG. 21 is a diagram showing another configuration example for realizing a digital resonance unit.

FIG. 22 is a diagram showing another configuration example for realizing a digital resonance unit.

FIG. 23 is a diagram showing another configuration example for realizing a digital resonance unit.

FIG. 24 shows a simulated impulse response of a guitar body resonating at 100 Hz.

FIG. 25 is a diagram showing the longest ringing mode (the least dumped component) of the impulse response shown in FIG. 24;

26 shows the initial impulse response of FIG. 24 with the least dumped components of FIG. 25 removed.

FIG. 27 is a diagram showing an impulse response of a guitar body calculated from data measured from an actual guitar.

FIG. 28 is a diagram showing parametric estimation values of two longest ringing modes in the impulse response shown in FIG. 27;

FIG. 29 shows the impulse response of FIG. 27 with the parametric components of FIG. 28 removed using inverse filtering.

FIG. 30 is a block diagram showing a tone synthesis system in which a delay loop with a filter is configured by a movable interpolation tap to synthesize a self-flanged string or a virtually detuned string.

FIG. 31 is a block diagram showing an embodiment of a string coupler simulation technique.

[Explanation of symbols]

 20 Table 26 Delay line 28 Loop filter 32 String 34 Resonator

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Julius O. Smith, The Sard United States 94306 California, Palo Alto, Cornell Street 2052 (56) References JP-A-53-88715 (JP, A) JP-A-4-52697 (JP, A) JP-A-5-11751 (JP, A) JP-A-4-204599 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10H7 / 08

Claims (43)

(57) [Claims] 1. A tone synthesis system for synthesizing a tone generated by the cooperative action of a vibration element and a resonance member acoustically coupled to the vibration element, comprising: an input unit for receiving an excitation signal ; Delay section to delay the signal
When closed loop connecting the filter for filtering the signal
An output unit for extracting an output from the closed loop
Hints, closed loop means for delay in the closed loop corresponds to the pitch of a musical tone to be synthesized, exciting means for supplying an excitation signal to the input portion having a component corresponding to the response of the resonant member tone synthesis system, characterized by comprising and. 2. The method according to claim 1 , wherein said exciting means includes a first member of said resonance member.
An excitation signal having a component corresponding to a partial response
For the tone signal output from the output unit.
A resonance element for adding resonance according to the second partial response of the resonance member.
Vibration filtering means, said first partial response and
The overall response of the resonant member by a second partial response
And a tone having a characteristic according to this response is synthesized.
2. The tone synthesis system according to claim 1, wherein: 3. The excitation means includes at least one table storing a value corresponding to the response of the resonance member, and a trigger means for reading out the value stored in the table and starting generation of a musical tone. Item 2. The tone synthesis system according to Item 1. Wherein said excitation means, interpolating a plurality of tables storing a value corresponding to at least one of said <br/> response, the values stored in the table based on at least one performance parameter tone synthesis system according to any one of claims 1 to 3, further comprising a means for providing said excitation signal by. Wherein said excitation means, according to claim 2 comprising a table means for storing data corresponding to the first partial response of said resonant member, and a reading means for reading the data stored from said table means The tone synthesis system according to item 1. 6. The tone synthesis system according to claim 5 , wherein said reading means reads said stored data from said table means in response to a player's operation. 7. A tone synthesis system according to any one of claims 1 to 6 wherein the excitation signal has a damped oscillation waveform. Wherein said excitation signal, tone synthesis system according to any one of claims 1 to 7 having a longer duration than one period of the musical tone to be synthesized. 9. The tone synthesis system according to claim 1, wherein the excitation signal is based on a record obtained by exciting a physical object corresponding to the resonance member. 10. The excitation means includes table means for storing data indicating a plurality of excitation signals, and the data stored in the table means stores a plurality of physical objects corresponding to a plurality of resonance members. 10. The tone synthesis system according to claim 9 , wherein the excitation means is obtained by excitation, and the excitation means further comprises a reading means for reading stored data corresponding to the selected excitation signal from the table means. 11. The tone synthesis system according to claim 1, wherein said excitation means includes a calculation means for calculating said excitation signal. 12. The musical sound synthesis system according to claim 11 , wherein said calculating means includes a backward filter for filtering a desired musical sound, and said backward filter includes a filter coefficient determined based on said vibration element. 13. The tone synthesis system according to claim 2 , further comprising a separating unit that separates the overall response into the first partial response and the second partial response. 14. The separation means according to claim 1 ,
14. The tone synthesis system according to claim 13 , including means for measuring a spectral peak in a Fourier transform to determine the second partial response in the tone synthesis. 15. The tone synthesis system according to claim 13 , wherein said separating means includes weighting means for determining said second partial response in said overall response using a weighting function and a filter design algorithm. . 16. The vibration element according to claim 1, wherein the vibration element is a string.
16. The musical sound synthesis system according to any one of 15 . 17. tone synthesis system according to any one of claims 1 to 15 wherein the resonant member is a body of the guitar. 18. The tone synthesis system according to any one of claims 1 to 15 wherein the resonant member is a body of the violin. 19. The tone synthesis system according to any one of claims 1 to 15 wherein the resonant member is a resonant plate and enclosure piano. 20. A tone synthesis system according to any one of claims 1 to 15 wherein the resonant member is made of a natural musical instrument. 21. The closed loop, wherein at least one additional closed loop is connected to the closed loop means.
Being adapted to be excited in the same manner and means, prior Symbol closed loop means and said at least one additional closed loop, tone according to any one of claims 1 to 20 in which a delay amount corresponding to a different tuning Synthetic system. 22. Before SL tone synthesis system of claim 21 further comprising a adding means for adding the output from the output and the at least one additional closed loop from the closed loop means. 23. the excitation means, also claim 1 including a table storing a value to be read in response to a trigger signal
Is a tone synthesis system according to 2. 24. The excitation signal, tone synthesis system according to claim 1 or 2 having a damped oscillation forms. 25. the excitation means, the response of the resonant member
A signal obtained by changing a signal consisting of components corresponding to
2. The tone synthesis system according to claim 1 , wherein the tone signal is supplied to the input unit as a signal . 26. The tone synthesis system according to claim 23 , wherein the table is repeatedly read to generate a sustained tone. 27. The excitation means has at least a part of a component corresponding to an impulse response of the resonance member.
Supplying a signal for the input portion as the excitation signal
Tone synthesis system according to claim 1 is intended. 28. The tone synthesis system according to claim 27 , wherein the excitation signal is a convolution result of the partial impulse response and an arbitrary excitation function. 29. The tone synthesis system according to claim 2 , wherein the convolution of the first partial response and the second partial response indicates an overall response of the resonance member . 30. The excitation means for supplying an excitation signal having a first damped oscillation form to the input section , and further comprising a second damping means which attenuates at a lower rate than the first damped oscillation form. A resonance adding means for adding resonance having a vibration form to the output of the output unit.
Item 2. The tone synthesis system according to Item 1 . 31. The excitation means for supplying a basic excitation signal, means for delaying the basic excitation signal by a predetermined amount, and the basic excitation signal and the delayed basic excitation signal adding the signals, tone synthesis system according to claim 30 comprising means for providing the addition result to said closed loop means. 32. A second excitation means for supplying a second excitation signal in parallel with the excitation signal, adding the excitation signal and the second excitation signal, and generating an excitation signal as a result of the addition. 31. The musical sound synthesis system according to claim 30 , further comprising: means for providing to said closed loop means . 33. A second excitation means for supplying a second excitation signal in parallel with the excitation signal, interpolating the excitation signal and the second excitation signal, and interpolating the interpolated excitation signal into the closed loop. And means for interpolating the means.
30. The tone synthesis system according to 30 . 34. the interpolation means, according to claim 3 including means for variably interpolate the excitation signal in response between the said second excitation signal to the control signal, and means for supplying said control signal
3. The tone synthesis system according to 3 . 35. The tone synthesis system according to claim 31 , wherein said closed loop means corresponds to a plucked string, and said predetermined amount of delay represents a position at which said string is plucked. 36. A combined output signal is generated by summing an output of said closed loop means and said second tone signal, and a second closed loop for circulating an excitation signal to provide a second tone signal. And means for performing
30. The tone synthesis system according to 30 . 37. The tone synthesis system according to claim 30 , wherein said resonance adding means includes means for generating a plurality of output signals, and further comprising means for adding each of said plurality of output signals. 38. The closed loop means, wherein the closed loop, claim 30 including the movable delay line taps to retrieve a second output which is delayed by a variable amount relative to the output
The tone synthesis system according to item 1. 39. The closed loop means, wherein the closed loop, tone synthesis system according to claim 30 comprising <br/> delay line taps fixed to retrieve a second output which is delayed with respect to the output . 40. The apparatus according to claim 40, further comprising a coupling filter for filtering the combined output signal, wherein the coupling filter comprises the closed loop means and the second closed filter.
37. The tone synthesis system of claim 36 , wherein the system generates a feedback signal supplied to at least one of the loops . 41. the excitation means, characterized in that it comprises a mixing means for providing a synthesized excitation signal by mixing a plurality of excitation tables, each storing different excitation signal, the output of said plurality of excitation table Claim 1 or
3. The tone synthesis system according to item 2 . 42. The tone synthesis system according to claim 41 , wherein said mixing means includes means for changing an amount of output of said plurality of excitation tables according to respective weighting factors. 43. The tone synthesis system according to claim 42 , further comprising means for changing said weighting function over time.