JP3237226B2 - Loop waveform generation device and loop waveform generation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a loop waveform generator and a loop waveform generator for generating a loop waveform from digital waveform data.

[0002]

2. Description of the Related Art An analog tone signal obtained by collecting the performance sound of an acoustic instrument such as a grand piano with a microphone or the like is converted from analog to digital. A sound source system based on a PCM (Pulse Code Modulation) system as a sound source has become widespread.

[0003] The PCM tone generator has the advantage that a tone that is extremely faithful to the sound of the original musical instrument can be produced, but in order to obtain a tone with good sound quality, the sampling frequency at the time of analog / digital conversion must be increased. It is necessary to increase the number of quantization bits. However, in this case, the data volume of the digital musical sound data becomes enormous, which causes an increase in the memory capacity and a significant increase in the cost of the electronic musical instrument.

To solve such a problem, there is a technique called a so-called loop waveform method. In the loop waveform method, digital tone data of one tone unit from the start of tone generation to silence is generally composed of an attack sound section (or a consonant section) having a large temporal change in frequency characteristics and poor periodicity; Paying attention to the fact that it is composed of a continuous sound section (or a vowel section) in which a change in frequency characteristics is small in time and a constant waveform is repeated in synchronization with the pitch cycle after that section, the digital tone in the attack sound section The data is stored as it is in the memory as sound source data, and only digital tone data of one repetitive waveform unit (hereinafter, referred to as a loop waveform) of a repetitive waveform constituting the interval is stored as digital tone data of a continuous tone section. It is stored in the memory as sound source data. At the time of musical tone reproduction, first, digital musical tone data corresponding to the attack tone section is read from the memory as it is and D / A converted, and thereafter, the loop waveform data corresponding to the sustain tone section is outputted until the mute instruction is given. The data is repeatedly read from the memory and D / A converted. Such a reproduction method will be referred to as loop reproduction.

In the above-described loop waveform method, it is only necessary to store digital musical tone data corresponding to an attack sound section and a loop waveform representing a sustained sound section in a memory, so that the memory capacity is greatly reduced. be able to.

[0006]

In the loop waveform method,
It is necessary to cut out a loop waveform representing the sustained sound section together with the digital sound data of the attack sound section from the original digital sound data of one sounding unit.

When a loop waveform is cut out, a first sampling point (hereinafter referred to as a loop start point) and a last sampling point (hereinafter referred to as a loop end point) of the digital musical sound data are determined. There is a need.

At the time of loop reproduction, each sampling point is repeatedly reproduced in a loop section from the loop start point to the sampling point one sample before the loop end point. In this case,
The amplitude values need to be connected smoothly at the connection points of the waveforms. Therefore, the loop start point and the loop end point need to be at the time when the amplitude is zero, that is, when the waveform crosses zero.

However, in general, since the sampling point does not always coincide with the position where the waveform crosses zero, conventionally, the loop start point and the loop end point could not completely coincide with the zero cross position of the waveform.

Therefore, conventionally, as shown in FIG. 24A, a sampling point close to the zero-cross position of the waveform has to be selected as a loop start point and a loop end point. As shown in the rectangular part, so-called loop noise occurs,
There is a problem that the tone quality of the reproduced electronic musical instrument is degraded.

The object of the present invention is based on a simple specification.
An object of the present invention is to make it possible to generate a loop waveform from digital waveform data such that a loop start point and a loop end point coincide with the zero-cross position of the waveform.

[0012]

According to the present invention, there is provided a digital waveform corresponding to one repetitive waveform unit to be repeatedly reproduced (loop reproduced) at a first sampling period from digital waveform data sampled at a first sampling period. It is assumed that a loop waveform generation device generates data as loop waveform data.

First, there is provided a pitch specifying means for specifying an arbitrary pitch at the time of repetitive reproduction (at the time of loop reproduction). Next, with respect to a predetermined period which is an integral multiple of one basic period corresponding to the pitch specified by the pitch specifying means and which is divisible within a predetermined allowable error range by the first sampling period,
There is provided sampling point number calculation means for calculating a number obtained by dividing the predetermined period by the first sampling period within a predetermined allowable error range as the number of sampling points.

Furthermore, based on the digital waveform data, prior to
From a first zero cross position is the head of the waveform segment of a predetermined period of predicates, the loop section extracting means for extracting a section up to the second zero-cross position that is the end of the waveform segment of the predetermined period, as a loop segment Have. For example, the loop section extracting means may first provide the user with a sampling point immediately before the first zero-cross position, which is the beginning of a waveform section for one repetitive waveform unit, and a sampling point at the end of the waveform section. The sampling points immediately before the zero-cross position 2 are designated as a loop start initial point and a loop end initial point, respectively. Next, the loop section extracting means calculates the time from the initial point of the loop start to the first zero-cross position immediately thereafter, shifts the entire waveform by that time, and matches the initial point of the loop start with the first zero-cross position. Then, the time from the loop end initial point corrected so as to be located immediately before the second zero-cross position to the second zero-cross position immediately thereafter is calculated, and the second zero-cross position is determined.

The digital waveform data of the above-described loop section is divided into loop section lengths such that the first zero cross position at the beginning of the loop section and the second zero cross position at the end of the loop section are included as sampling points. Is re-sampled at the second sampling period obtained by dividing by the above-mentioned number of sampling points, thereby generating loop waveform data. The re-sampling process by this means is, for example, a method in which a sample value at each time position that advances in the second sampling period is obtained by sampling the sampling points of the digital waveform data sampled in the first sampling period near the above-described time position This is a process of calculating by an interpolation operation using a Lagrange's interpolation function or the like.

The present invention can also be realized as a loop waveform generation method for executing each step corresponding to the function of each of the above-described means.

[0017]

According to the present invention, the sampling period ( sample
If the readout cycle is constant, the fact that the number of sampling points per basic cycle of a musical tone and the pitch at which the musical tone is reproduced corresponds to one to one is used.

In this case, one basic cycle corresponding to an arbitrary pitch specified by a user or the like is not always divisible by a fixed sampling cycle. However, if the cycle is an integral multiple of one basic cycle corresponding to a designated arbitrary pitch, the cycle of the integral multiple is almost always divisible by a fixed sampling cycle.

Therefore, according to the present invention, first, a predetermined period which is an integral multiple of one basic period corresponding to an arbitrary pitch specified by a user or the like and is within a predetermined allowable error range by the first sampling period. A waveform signal having a designated pitch is sampled at the first sampling period, and a number obtained by dividing the predetermined period within the predetermined allowable error range by the first sampling period with respect to such a divisible period is sampled. This is calculated as the number of sampling points per predetermined period of the waveform signal in that case.

Next, when a waveform section for the above-mentioned predetermined period to be loop-reproduced is designated on the digital waveform data based on, for example, a designation by a user, the first zero crossing which is the head of the waveform section is performed. A loop section from the position to the second zero-cross position at the end of the loop section is extracted.

The length of the loop section is calculated such that the above-mentioned loop section includes the first zero-cross position at the beginning of the loop section and the second zero-cross position at the end of the loop section as sampling points. Is resampled in the second sampling period obtained by dividing by the “number of sampling points”.

As a result, loop waveform data composed of “calculated number of sampling points + 1” samples is generated, and the loop start point as the start point and the loop end point as the end point are set at the beginning of the loop section. It exactly matches the first zero-cross position and the last second zero-cross position.

The loop waveform data for the “calculated number of sampling points + 1” samples thus generated is saved in a memory or the like. In the saved loop waveform data, the sampling point from the sampling point at the first zero-cross position at the beginning of the loop section to the sampling point one sample before the sampling point at the second zero-cross position at the end of the loop section,
Loop waveform data composed of “calculated sampling point number” samples is loop-reproduced in the first sampling cycle, so that the loop waveform data does not include loop noise and has no auditory problem within a specified range. A waveform almost matching the pitch is obtained.

[0024]

Embodiments of the present invention will be described below in detail with reference to the drawings. <Structure of Embodiment of the Present Invention> FIG. 1 is an external view of an embodiment of the present invention, and FIG.

When the control / processing unit 205 starts a control process for generating a loop waveform, first, PCM musical tone waveform data for one tone unit as a source is stored in a hard disk or a hard disk installed in the waveform input / output unit 206. The source waveform file 101 stored on the floppy disk is loaded into the RAM 204 in the personal computer 103, and the waveform is displayed on the display 102 of the display unit 202.

Next, the user specifies the start and end positions of the waveform section to be looped on the displayed waveform by using the mouse 105 as a pointing device. When this designation is made, the control / processing unit 205 starts arithmetic processing for generating a loop waveform, and as a result,
A desired loop waveform is generated, and is output to the waveform input / output unit 20.
6 is stored in a created waveform file 106 stored on a hard disk or a floppy disk installed in the computer 6.

The ROM 203 of the control / processing unit 205 includes
A series of control programs related to the loop waveform generation processing are stored, and the RAM 204 stores the above-described PCM musical tone waveform data and various variables used in the loop waveform generation processing. <First Operation Principle of the Present Invention> As the first operation principle of the present invention having the above-described configuration, an operation principle when loop waveform data having a predetermined reference pitch is generated will be described.

The first principle of operation is that if the sampling period at the time of reproduction is constant, the number of sampling points per basic period of a musical tone and the pitch at which the musical tone is reproduced are one.
It takes advantage of the fact that there is a one-to-one correspondence.

First, the user determines a reference pitch. Next, the user determines a sampling period T that divides one basic period corresponding to the reference pitch.

Next, when a musical tone having a reference pitch is sampled at the sampling period T, the number M of sampling points per basic period of the musical tone is calculated by the following equation.

[0031]

(Equation 1)

Subsequently, the analog tone signal having an arbitrary pitch is A / D-converted at the sampling period T to obtain PCM tone waveform data. Next, the user makes a continuous sound section of the PCM musical tone waveform data obtained by the A / D conversion,
A waveform section for one basic cycle starting from the zero-cross position and ending at the zero-cross position is registered, and the start and end of the waveform section are designated.

Based on this designation, the sampling point closest to the designated start position and immediately after the zero-cross position of the waveform, and the sampling point closest to the designated end position and immediately following the zero-cross position of the waveform Sampling points to be calculated are calculated as a loop start initial point and a loop end initial point, respectively.

Based on the calculated loop start initial point and loop end initial point, the loop end initial point is calculated from the zero cross position at the beginning of the above-described one basic period waveform section immediately after the loop start initial point. The section up to the zero-cross position at the end of the above-described one basic cycle of the waveform section located immediately after is extracted as a loop section, and the loop section length is calculated.

Then, at the sampling period T 'calculated by the following equation (2), the above loop section is resampled (resampled) so that the leading zero-cross position and the trailing zero-cross position are included as sampling points. ) Is done.

[0036]

(Equation 2)

As a result, loop waveform data including M + 1 samples including the sampling points at the zero cross positions at both ends of the loop section is generated, and the loop start point as the start point and the loop end point as the end point are represented by the waveform Exactly matches the leading zero-cross position and the trailing zero-cross position.

The loop waveform data for M + 1 samples thus generated is saved in a memory or the like.
Of the saved loop waveform data for M + 1 samples, a loop consisting of M samples from the sampling point at the beginning zero-cross position of the loop section to the sampling point one sample before the sampling point at the end zero-cross position of the loop section. By performing loop reproduction of the waveform data at the sampling period T,
A tone that does not include loop noise and that exactly matches the reference pitch can be obtained. <Specific Operation of Embodiment of the Present Invention (Part 1)> A specific operation of the embodiment of the present invention for generating loop waveform data based on the above-described first operation principle of the present invention will be described.

First, the user determines, for example, the pitch of A4 whose fundamental frequency is 442.000 Hz as the reference pitch. Next, the user sets, for example, A4 as a sampling period T that divides one basic period corresponding to the reference pitch.
A sampling period T = 1/5304 [seconds] that divides 1/442 [second], which is one basic period corresponding to the pitch of sound, is determined. The user operates the key input unit 201 shown in FIG.
Set to 06. The sampling frequency in this case is 5
304 [Hz].

Next, the control / processing unit 205 shown in FIG. 2 performs one of the musical tones having the reference pitch at the sampling period T on the basis of the above equation (1).
The number M of sampling points per basic cycle is calculated. For example, when a musical tone having a pitch of A4 is sampled at a sampling period T = 1/5304 [seconds], the number M of sampling points per basic period of the musical tone is calculated as follows from Expression 1. Is done. M = (1/442) / (1/5304) = 12 [points] The number M of sampling points per one basic period corresponding to the reference pitch calculated as described above is RA in FIG.
It is stored in M204.

Subsequently, an analog tone signal having an arbitrary pitch is A / D-converted at a sampling period T based on a user's operation on the key input section 201, and PCM tone waveform data is converted into a waveform input / output signal shown in FIG. The source waveform file 101 is stored on a hard disk or a floppy disk installed in the unit 206.

Thereafter, based on the user's operation on the key input unit 201, the control / processing unit 205 starts a control processing program for generating a loop waveform shown by the operation flowchart of FIG.

First, in step S301, PCM musical tone waveform data for one tone unit as a source is obtained from a source waveform file 101 stored in a hard disk or a floppy disk installed in the waveform input / output unit 206 and stored in the personal computer main unit 103. While being loaded into the RAM 204, the waveform is displayed on the display 102 of the display unit 202.

Next, when the user designates the beginning (loop start) and end (loop end) of the waveform section to be loop-reproduced on the displayed waveform using the mouse 105, the determination in step S302 becomes YES. .

In the next step S303, the sampling point closest to the specified head position and immediately after the zero cross position of the waveform, and the sampling point closest to the specified end position and immediately after the zero cross position of the waveform The sampling points to be calculated are calculated as a loop start initial point LSIP and a loop end initial point LEIP, respectively, as shown in FIG.
These points are stored in the PC read into the RAM 204.
The leading sampling point of the M musical tone waveform data is numbered “0”, and is represented by numbers (0, 1, 2, 3,...) That sequentially increase for each subsequent sampling point.

In the subsequent steps S304 to S308, based on the above-described calculated LSIP and LEIP, the position of the zero-cross position at the head of the waveform section of one basic period immediately after the LSIP is shifted to the position immediately after the LEIP. The section up to the zero-cross position at the end of the above-described waveform section for one cycle is extracted as a loop section.

First, in step S304, as shown in FIG. 4A, the loop start initial point LSIP
Is calculated from the time Δt _LS to the zero-cross position immediately after.

The detailed operation of step S304 for calculating the above-mentioned time Δt _LS will be described with reference to FIGS.
A description will be given based on the operation flowchart of FIG. In FIG. 10, first, in step S1001, the RAM2
The base time point b stored as a variable in 04 is initialized to the time point LSIP · T of the loop start initial point LSIP. Here, T is a sampling period when the waveform data stored in the RAM 204 is A / D converted.

Next, in step S1002, the RAM 20
The value 2 is substituted for the variable m stored in 4 and the time d stored as a variable in the RAM 204 is set to the value T / m in the subsequent step S1003. Therefore, d = T / 2.

Subsequently, at step S1004, a time point x = b + d, which is a time d = T / 2 from the base time point b initially set at the loop start initial point LSIP, is calculated. The time point x represents the elapsed time from the beginning of the musical tone waveform stored in the RAM 204, and x is the RAM time.
This is a variable stored in 204.

Next, in step S1005, the interpolation value f (x) of the musical tone waveform at the time point x is calculated using the Lagrange interpolation formula described later. This interpolation value f (x)
Are also variables stored in the RAM 204.

In step S1006, the interpolation value f (x)
Is determined to be "0" on the computer calculation accuracy. If this determination is NO, step S1007
, The sign of the interpolation value f (x) and the sample value SMP (L
It is determined whether the sign of (SIP) is the same.

Now, for example, as shown in FIG. 5, the zero-cross position of the tone waveform is determined by the loop start initial point L.
SIP and the next sampling point (LSIP +
Located between 1).

Then, as shown in FIG. 5, the sample value SMP at the loop start initial point LSIP is obtained.
When (LSIP) and the interpolation value f (x) at the time point x are the same, the zero cross position exists at a time point after the time point x = b + d.

In this case, in step S1008, as shown in FIG. 6, the base time point b is shifted to the time point LSIP · T + T / 2 after the time d = T / 2 has elapsed from the original base time point LSIP · T. It is. Then, in step S1009, the value 2 of the variable m is doubled to a value 4, and in step S1003, the time d is reset to a new time T / m. In FIG. 6, the time d = is changed from the original time T / 2 to a new time T /
It is reset to 4. Then, in step S1004, a time x = b + after a new time d = T / m = T / 4 has elapsed from the new base time b = LSIP · T + T / 2.
d is calculated, and in the next step S1005, the interpolation value f
(X) is calculated. After the determination in step S1006, the sign of the interpolation value f (x) is determined in step S1007.

In the case of FIG. 6, the sample value SMP (LS
IP) and f (x) are opposite signs, and in this case, the zero cross position exists at a time point before the time point x. In this case, since step S1008 is not executed, as shown in FIG. 7, the base time point b = LSIP
• T + T / 2 is not moved. Then, step S10
In 09, the value 4 of the variable m is doubled to a value 8,
In step S1003, the time d is changed to a new time T /
m. In FIG. 7, the time d is equal to the original time T /
From 4, the time is reset to a time T / 8 corresponding to a half thereof. Then, in step S1004, a new time d = T / m = from the base time point b = LSIP · T + T / 2.
The time point x = b + d after a lapse of T / 8 is calculated, and in the next step S1005, the interpolation value f (x) is calculated. After the determination in step S1006, step S10
At 07, the sign of the interpolated value f (x) is determined.

In the case of FIG. 7, the sample value SMP (LS
IP) and f (x) are opposite signs, and in this case, the zero cross position exists at a time point before the time point x. In this case, as in the case of FIG.
8, the base time b = LSIP.T + T / 2 is not moved, as shown in FIG. Then, in step S1009, the value 8 of the variable m is doubled to a value 16, and in step S1003, the time d is reset to a new time T / m. In FIG. 8, the time d is reset from the original time T / 8 to a time T / 16 corresponding to a half thereof. Then, step S100
At 4, a new time d = T / m = time T / 16 after the base time b = LSIP · T + T / 2, x =
b + d is calculated, and in the next step S1005, an interpolation value f (x) is calculated. After the determination in step S1006, the sign of the interpolation value f (x) is determined in step S1007.

In the case of FIG. 8, the sample value SMP (LS
IP) and f (x) have the same sign, and in this case, the zero cross position exists at a time point after the time point x. In this case, as in the case of FIG.
At 08, as shown in FIG.
Is the time d = from the original base time LSIP · T + T / 2
LSIP · T + T / 2 + T /
Moved to 16. Thereafter, in step S1009, the value 16 of the variable m is doubled to a value 32, and in step S1003, the time d is reset to a new time T / m. In FIG. 9, the time d is reset from the original time T / 16 to a time T / 32 corresponding to a half thereof. Then, in step S1004, a new time d from the base time point b = LSIP · T + T / 2 + T / 16
= T / m = T / 32, x = b + d is calculated, and in the next step S1005, an interpolation value f (x) is calculated. After the determination in step S1006, the sign of the interpolation value f (x) is determined in step S1007.

As described above, as the base time point b is sequentially shifted and the time d added thereto is reduced, the interpolation value f (x) of the waveform signal gradually decreases, and finally, the calculation accuracy of the computer is reduced. It becomes almost equal to the above “0”, and the determination in step S1006 becomes YES. The time point x at this time is the time point of the zero cross position immediately after the loop start initial point LSIP.

Therefore, the loop start initial point LS
Time Δt from IP to zero cross position immediately after IP
_In step S1010, _{LS determines} that Δt _LS = x−LS
It can be obtained as IP · T.

As described above, the loop start initial point LSI
FIG. 10 for calculating an interpolation value f (x) at time point x in the operation of step S304 in FIG. 3 shown in the operation flowchart of FIG. 10 for calculating the time Δt _LS from P to the zero-cross position immediately thereafter. Step S1
The calculation of 005 will be described.

The interpolation value f (x) is calculated using Lagrange's interpolation formula expressed by the following equation (3).

[0063]

(Equation 3)

The Lagrange interpolation formula will be described below. Figure 11 shows the linear interpolation to be interpolated official basis Lagrange, original waveform f = the sampling point when x ₀ and x ₁ of f (x), each sample value f (x ₀₎ and f From (x ₁ ), an equation of a straight line (primary equation) passing through two points is obtained, and by substituting x into the equation of the straight line, the interpolation value f (x) at the time point x is calculated.

On the other hand, as shown in FIG.
= Sampling point times x ₀ , x ₁ on f (x),
And a x _2, each of the sample values _{f (x 0), f (} x
₁ ) and f (x ₂ ), an equation of a quadratic curve shown by a thin line in FIG. 12 is obtained, and by substituting x into the equation, the original waveform can be more faithfully compared to the above-described linear interpolation. Time x
Is calculated.

As described above, the sample value on the original waveform is
If the number of points is increased to three, four,..., N, interpolation data with higher accuracy can be obtained. Lagrange's interpolation formula is an interpolation formula using sample values of n points derived based on such a concept.
It is shown by the formula.

In the equation (3), x is an interpolation position to be obtained.
f (x) is an interpolation value to be obtained. X ₀ ,
x ₁ ,..., x _n are the sampling points of the original waveform, and f (x ₀ ), f (x ₁ ),.
_n ) indicates the sample value at each sampling point. If n = 1 is set in this equation, the equation becomes a linear interpolation equation. That is, linear interpolation can be said to be two-point Lagrange interpolation.

FIG. 13 shows step S in FIG.
In 1005, when the interpolation value f (x) at the time point x is calculated, and when n = 15 in the above equation 3, the sampling point time points x ₀ , x ₁ ,.
₁₅ and sample values f (x ₀ ), f (x ₁ ),.
The value to be substituted for (x ₁₅ ) is shown.

[0069] First of all, the time LSIP · T of the loop start initial point LSIP to the sampling point point x ₇
There is assigned, also from x ₀ from x ₇ 7 temae, to 16 points to x ₁₅ in the x ₇ after eight points, each sampling point time (LSIP-7) from · T (LSIP +
8) Up to T are substituted.

Next, the sample values f (x ₀ ) to f (x ₁₅ )
Are the sampling points (LSIP-7), (L
, (LSIP + 8), SMP (LSIP-7), SMP (L
SIP-6),... SMP (LSIP + 8) are substituted. Note that 0 is substituted for the sample value corresponding to the sampling point at which the value becomes a negative value.

Equation 3 is calculated as described above, and FIG.
The interpolation value f (x) at the time point x obtained in step S1004 is obtained. As described above, in step S304 of FIG. 3, as shown in FIG. 4A, the time Δt _LS from the loop start initial point LSIP to the immediately succeeding zero-cross position is calculated.

Next, in step S305 in FIG.
As shown in FIG. 4B, with respect to the PCM musical tone waveform data A for one sounding unit stored in the RAM 204,
The time points of all sampling points are shifted in a direction delayed by the above-mentioned time Δt _LS, and as a result, FIG.
Waveform data as shown by B in FIG.

The processing in step S305 in FIG. 3 will be described in more detail with reference to the operation flowchart in FIG. First, in step S1401, a sampling point k indicating the number of each sampling point from the beginning of the waveform data is initialized to zero. This point k is stored in the RAM 204 as a variable.

Next, the PCM tone waveform data, the interpolation value at the time delayed by a time Delta] t _LS from time kT of the sampling point k is, as f (kT + Δt _LS) based officially interpolation Lagrange equation (3) described above Is calculated, and the interpolation value f (kT + Δt _LS ) is calculated at the sampling point k.
RAM 204 as a new sample value SMT ′ (k)
Is stored.

FIG. 15 shows the interpolation value f (kT + Δt)
_LS ) is calculated, and when n = 15 in the above equation (3), sampling point times x ₀ , x ₁ ,.
..., and x _15, the sample value _{f (x 0), f (} x 1), ·
.., The values to be substituted for f (x ₁₅ ) are shown.

First, the time point kT of the sampling point k is substituted for the sampling point time point x ₇ , and x
_For 16 points from x ₀ 7 points before ₇ to x ₁₅ 8 points after x ₇ , each sampling point time point (k−
7) · T to (k + 8) · T are respectively substituted.

Next, the sample values f (x ₀ ) to f (x ₁₅ )
Has sampling points (k-7), (k-
6),..., (K + 8), sample values SMP (k−7), SMP (k−6),... Of the original PCM musical tone waveform data stored in the RAM 204.
SMP (k + 8) is substituted. Note that 0 is substituted for the sample value corresponding to the sampling point at which the value becomes a negative value.

Equation 3 is calculated in this way, an interpolation value f (kT + Δt _LS ) is calculated, and a new sample value SM is calculated.
P ′ (k). Subsequently, in step S1403, the value of the sampling point k is advanced by one, and in the next step S1404, the value of the k is stored in the RAM 204.
Is determined to have exceeded the waveform end point WEP, which is the last sampling point of the original PCM musical tone waveform data stored in.

If the determination is NO, step S
Returning to 1402, the interpolation value at the time point delayed by the time Δt _LS from the time point kT of the new sampling point k is calculated as f (kT +
Δt _LS ), and the interpolation value f (kT + Δ)
t _LS ) is the new sample value S at sampling point k
It is stored in the RAM 204 as MP ′ (k).

When the sampling point k exceeds the waveform end point WEP, the processing ends. In step S305 of FIG.
Of the PCM musical tone waveform data for one sounding unit stored in No.4, the time points of all sampling points correspond to the time Δt.
Shifted in the direction delayed by _LS .

As a result of the processing in step S305 in FIG. 3 described above, as shown by B in FIG. 4B, in the waveform data shifted in the direction delayed by the time Δt _LS , the initial loop start The sampling point corresponding to the point LSIP exactly matches the zero-cross position of the musical sound waveform.

On the other hand, in the original PCM tone waveform data, the loop end initial point LEIP was the sampling point immediately before the zero cross position of the waveform, but Δ
In new waveform data obtained by shifting the waveform in the direction delayed by the time t _LS , the sampling point corresponding to the loop end initial point LEIP may be shifted immediately after the zero-cross position of the waveform. That is, Δt _LS is a time shorter than the sampling period T when the original tone waveform data is A / D converted, and
Loop end initial point LEIP on original tone waveform
Is also shorter than the sampling period T. Therefore, in the new waveform data obtained by shifting the waveform in the direction delayed by the time Δt _LS according to the magnitude relationship between the two, the sampling point corresponding to the loop end initial point LEIP remains immediately before the zero-cross position of the waveform. , Or just move right after.

Therefore, in step S306 of FIG. 3, the sampling point immediately before the zero-cross position at the end of the loop section on the new musical tone waveform obtained by shifting the waveform is changed to the corrected loop end initial point LE.
Correction processing for re-determining as IP 'is executed.

Now, the sample value SMP (LE) of the loop end initial point LEIP on the original musical sound waveform data
IP) and the original loop end initial point L on new tone waveform data obtained by shifting the waveform.
Sample value S of sampling point corresponding to EIP
The sign of MP '(LEIP) is determined.

If these codes are the same, the sampling point corresponding to the original loop end initial point LEIP on the new musical tone waveform data obtained by shifting the waveform remains immediately before the zero-cross position of the waveform. Therefore, the LEIP is directly used as the corrected loop end initial point LEIP ′.

On the other hand, if the two codes are different, new tone waveform data obtained by shifting the waveform
Since the sampling point corresponding to the original loop end initial point LEIP is shifted immediately after the zero-cross position of the waveform, the sampling point (LEIP-1) immediately before the sampling point corresponding to the original loop end initial point LEIP is , The corrected loop end initial point LEIP ′.

As described above, a corrected loop end initial point LEIP 'is newly obtained. next,
In step S307 in FIG. 3, as shown in FIG. 4B, the corrected loop end initial point LEI
The time Δt _LE from P ′ to the zero cross position immediately after P ′ is calculated.

The process of calculating the time Δt _LE described above is the same as the process of calculating the time Δt _LS from the loop start initial point LSIP to the zero cross position immediately after that in step S304 shown in the operation flowchart of FIG. However, the loop start initial point LSIP in step S1001 in FIG. 10 is replaced with a loop end initial point LEIP. Step S100
7, the sample value SMP (LSIP) of the original musical sound waveform data in the LSIP of FIG. 7 is replaced with the sample value SMP ′ (LEIP) of the waveform-shifted waveform data in the LEIP.
Further, Δt _LS in step S1010 is replaced with Δt _LE .

After the time Δt _LE is calculated by the above-described processing in step S307 in FIG. 3, next, in step S308 in FIG. 3, the loop start time (loop start time) LST and the loop end time (loop end time)
The calculation of LET is performed. The loop start time LST refers to the zero cross position at the beginning of the loop section of the waveform data, and the loop end time LET refers to the zero cross position at the end of the loop section of the waveform data.

Here, the loop start time LST is, as shown in the upper part of FIG. 16, the step S305 in FIG.
At the sampling point corresponding to the original loop start initial point LSIP in the waveform data newly obtained in the RAM 204 by the waveform shift processing. That is, the loop start time LST can be obtained by the following equation.

[0091]

(Equation 4)

Is as follows. However, T means that the original waveform data is A
This is a sampling cycle when the / D conversion is performed. on the other hand,
As shown in the upper part of FIG. 16, the loop end time LET is a time Δt _LE at the time of the corrected loop end initial point LEIP ′ obtained in step S306 of FIG.
Is equal to the time when That is, the loop end time LE
T can be obtained by the following equation.

[0093]

(Equation 5)

According to the above equations (4) and (5), in step S308 in FIG.
And the loop end time LET is calculated, and in step S309 of FIG. 3, the first operation principle of the present invention is described above.
A sampling period T ′ for performing resampling including LST and LET as sampling points is calculated by the following equation corresponding to the equation (2).

[0095]

(Equation 6)

Here, M is one basic tone of a tone having a reference pitch (for example, A4) when the tone is sampled at the sampling period T, as described in <First principle of operation of the present invention>. This is the number of sampling points per cycle, and is calculated based on Equation 1 at the start of the loop waveform data generation process, as described above.
4 is stored in advance. For example, when a musical tone having a pitch of A4 is sampled at a sampling period T = 1/5304 [sec], M = 12 as described above.

Subsequently, in step S310 in FIG.
In the sampling period T ′ calculated based on the above equation 6, the loop section from the loop start point LST to the loop end point LET includes the leading zero-cross position and the trailing zero-cross position as sampling points. As a result,
Loop waveform data consisting of M + 1 samples including the sampling points of the zero cross positions at both ends of the loop section is generated. The loop start point as the start point and the loop end point as the end point are defined as the zero cross position at the beginning of the waveform. Exactly matches the trailing zero-cross position. M + 1 generated in this way
Of the loop waveform data for the sample, the loop period of the loop waveform data consisting of M samples from the sampling point at the leading zero-cross position of the loop section to the sampling point one sample before the sampling point at the zero-cross position at the end of the loop section is sampled. By performing loop reproduction at T, it is possible to obtain a musical tone that does not include loop noise and that exactly matches the reference pitch.

However, in the embodiment of the present invention, it is possible to reproduce the attack sound section before the loop reproduction of the loop section,
Further, if necessary, for example, to enable reproduction of a release section after loop reproduction of the loop section,
Of the above-mentioned waveform-shifted waveform data stored in M204, not only the loop section but also the entire waveform data is resampled.

In this case, when the entire waveform data is resampled at the sampling period T 'from the beginning of the waveform,
Normally, the loop start time LST in FIG. 16 does not coincide with the sampling point. This is because the time from the beginning of the waveform data to the loop start time LST is not always divisible by the sampling period T '.

Therefore, in order to match the loop start time LST with the sampling point, when the entire waveform data is resampled, the time at which resampling is started is determined by a predetermined time from the time of the first sampling point of the waveform data. Staggered by time.

Step S3 of FIG. 3 including such processing
The resampling process of No. 10 will be described with reference to the detailed operation flowchart of FIG. First, in step S1701, a sampling point k indicating the number of each sampling point from the beginning of the waveform data is initialized to zero. This point k is stored in the RAM 204 as a variable.

In step S1702, an initial value i at the time when resampling is started is calculated. This initial value i is equal to the time value obtained by subtracting the integer part INT (LST / T ') of the time value LST divided by the sampling period T' from the time value LST at the time of the loop start. That is, the initial value i is calculated by the following equation.

[0103]

(Equation 7)

In the subsequent repetition of the processing in steps S1703 to S1707, the waveform values at each time point starting from the initial value i and proceeding by the sampling period T 'are calculated based on the Lagrangian interpolation formula of the above equation (3). Are sequentially taken as sampling values, whereby a resampling process is executed.

Steps S1703 to S1707 in FIG.
Will be described with reference to the explanatory diagrams of FIGS. First, as shown in FIG. 18A, for example, assuming that the sampling cycle T = 3, in the waveform-shifted waveform data already stored in the RAM 204, each sampling point 0, 1, 2,. , 6,... Corresponding to the sample values SMP '(0), SMP'
(1), SMP '(2), ..., SMP' (6), ...
.. Indicates that each sampling time point x =
0, 3, 6,..., 18,.

Then, as shown in FIG. 18 (b), for example, assuming that the sampling period T '= 4, each sampling time i = 4 starts at a time 1 shifted from the sampling time 0 and is located at an interval of T'. 1, 5, 9, 13,
17, resampling is performed, and each sampling point k =
New sample values OUT (k) = OUT (0), OUT (1), OUT for 0, 1, 2, 3, 4,.
(2), OUT (3), OUT (4),... Are calculated.

At this time, the sampling value OUT (k) of the sampling point k corresponding to each sampling time point i located at the interval of T 'includes the sampling point T including the sampling time point i and eight points before and after the sampling point i, for a total of 16 points. Using the sample values of the waveform data in the RAM 204 at the sampling point corresponding to the sampling time point x located at the interval of, the calculation is performed based on the Lagrangian interpolation formula of the above equation (3).

Here, as shown in FIG. 19, of the sampling times x located at the above-mentioned interval T, T ′
The sampling point j corresponding to the sampling time point immediately before each sampling time point i located at the interval of is represented by an integer part INT (i / i / （) obtained by dividing the sampling time point i by the original sampling period T, as shown in the following equation. T).

[0109]

(Equation 8)

For example, in FIG. 18, for example, i = 13
Then, j = 4 is calculated. Therefore, the sampling point immediately before the sampling point i = 13 is j = 4, and the corresponding sampling point x is x = j · T =
4 × 3 = 12, and the sampling point j =
The sample value of the waveform data stored in the RAM 204 corresponding to No. 4 is SMP '(4).

The sampling point j described above corresponds to FIG.
7 is calculated in step S1703. Next, after the determination in step S1704 in FIG. 17 described later, in step S1705 in FIG. 17, as shown in FIG. 19, the interpolation value corresponding to the sampling time i is calculated by the Lagrange interpolation formula of the above-described equation (3). F (i), which is stored in the RAM 204 as a new sample value OUT (k) at the sampling point k.

[0112] Figure 20, when the interpolated value f (i) is calculated, when the n = 15 in Equation 3 described above, the sampling point point _{x 0, x 1, ···,} x 15 And sample values f (x ₀ ), f (x ₁ ),..., F (x ₁₅ )
The values to be assigned to are shown.

First, the time point j · T of the sampling point j is substituted for the sampling point time point x ₇ , and
From x ₀ from x ₇ 7 temae, x ₁₅ to from x ₇ after 8 points
For each of the 16 points up to the point of each sampling point (j
−7) · T to (j + 8) · T are substituted.

Next, the sample values f (x ₀ ) to f (x ₁₅ )
Has sampling points (j-7), (j-
6),..., (J + 8), sample values SMP ′ (j−7), SMP ′ (j−6) of the waveform-shifted waveform data stored in the RAM 204.
... SMP '(j + 8) is substituted. Note that 0 is substituted for the sample value corresponding to the sampling point at which the value becomes a negative value.

Thus, step S170 in FIG.
5, a new sample value OUT of one sampling point k corresponding to one sampling time i
After (k) is obtained, in step S1706 of FIG. 17, the sampling point k is advanced by one, and in step S1707, the sampling point i is advanced by the sampling period T '.

Thereafter, the flow returns to step S1703, where the sampling point j corresponding to the sampling point immediately before the new sampling point i is calculated. Then, in step S1704, the value of j is R
It is determined whether or not a waveform end point WEP which is a final sampling point of the waveform-shifted waveform data stored in the AM 204 has been exceeded.

If the determination is NO, step S1705
Then, a new sample value OUT (k) corresponding to a new sampling time point i is calculated, and when the above determination is YES, the resampling process of step S310 in FIG. 3 is ended.

By the resampling process described above, FIG.
As shown in FIG. 6, waveform data including loop waveform data including M + 1 samples including sampling points at the zero cross positions at both ends of the loop section is generated, and a loop start point LSP which is a start point of the loop section and an end point thereof. Is exactly the same as the loop start time LST, which is the beginning zero cross position of the loop section, and the loop end time LET, which is the end zero cross position.

Finally, in step S311 of FIG. 3, the above-mentioned loop start point LSP and loop end point LEP are set to the waveform data OUT obtained in the RAM 204 by the resampling process in step S310.
(K) is added as header data to the head portion, and in the next step S312, the waveform data including the above header data obtained in the RAM 204 is stored in the waveform input / output unit 2 shown in FIG.
The saved waveform file 106 is saved on the hard disk or floppy disk installed at 06 as the created waveform file 106 in FIG. 1, and the loop waveform generation processing ends.

The waveform data including the loop waveform data for M + 1 samples generated in this way is stored in the ROM of the electronic musical instrument later and used as a tone generator. In this case, the waveform data before the loop section is read out from the memory as it is at the sampling period T and reproduced, so that, for example, the attack sound section is reproduced, and then one sample before the loop end point LEP from the loop start point LSP. By performing loop reproduction of the loop waveform data composed of M samples up to the sampling point at the sampling period T, it is possible to generate a musical tone that does not include loop noise and that exactly matches the reference pitch. <Second Operation Principle of the Present Invention> As a second operation principle of the present invention, an operation principle when loop waveform data having an arbitrary pitch is generated will be described.

First, the user previously sets a reference pitch,
The point that a sampling period T that divides one basic period corresponding to the reference pitch is determined is the same as in the case of the first operation principle described above. In the second operating principle, the user further specifies an arbitrary pitch.

Next, according to the first principle of operation described above, when a musical tone having a reference pitch is sampled at a sampling period T, the number M of sampling points per basic period of the musical tone is calculated by the following equation (1). Was done. On the other hand, in the second operation principle, the following calculation is performed.

As described above, since one basic period corresponding to the reference pitch is divisible by the sampling period T, the musical tone having the reference pitch is calculated by the above-mentioned equation (1) in the first operation principle. The number M of sampling points per basic period of the musical tone when sampled at the sampling period T is an integer. Therefore, as described above, if the loop waveform data of one basic cycle consisting of M points is reproduced in a loop at the sampling cycle T, the pitch exactly matches the reference pitch.

On the other hand, one basic cycle corresponding to a designated arbitrary pitch is not necessarily the sampling cycle T described above.
Is not necessarily divisible by Therefore, even if the loop waveform data of one basic cycle consisting of integer points is loop-reproduced at the sampling cycle T, the pitch may not be able to match the specified pitch.

However, if the period is an integral multiple of one basic period corresponding to a designated arbitrary pitch, the period of the integral multiple is almost always divisible by the sampling period T. For example, as described above, the reference pitch is A4 = 442.
[Hz], and a sampling period T that divides one basic period corresponding to the pitch is set to 1/5304 [seconds]. Then, for example, the pitch of G4 is 393.777 [Hz], and one basic cycle corresponding to the pitch is not an integral multiple of the sampling cycle T as shown by the following equation. (1 / 393.777) / (1/5304) ≒ 13.47 However, twice of one basic period corresponding to the pitch of G4 is approximately 27 times the sampling period T as shown in the following equation.
Times, that is, an integral multiple. {(1 / 393.777) × 2} / (1/5304) {26.94} 27 The above relationship is shown in FIG.

Therefore, if the loop waveform data of the above-mentioned two basic cycles consisting of 27 points is reproduced in a loop at the sampling cycle T, the pitch is calculated as shown in the following equation, and the pitch is within a range where there is no problem in hearing. Within, the pitch of the designated G4 is almost equal to 393.777. 1 / [{(1/5304) × 27} / 2] = 392.889 [Hz] Therefore, first, a sampling period which is an integral multiple of one basic period corresponding to a designated arbitrary pitch, T defines a cycle that is divisible within a predetermined allowable error range. Then, a loop waveform corresponding to the integral multiple of the period consisting of a number of sampling points obtained by dividing the integral multiple of one basic cycle corresponding to the designated pitch by a sampling cycle T within a predetermined allowable error range. If the data is loop-reproduced at the sampling period T, the pitch substantially matches the designated pitch within a range where there is no problem in hearing.

That is, first, an integer n such that the value of the following equation becomes an integer within a predetermined allowable error range is determined. The tolerance is, for example, about ± 0.2.

[0128]

(Equation 9)

Next, using the above determined integer n, the integer closest to the result of the operation of Expression 9 is calculated by the following expression.

[0130]

(Equation 10)

The operation "INT" is an operation for extracting the integer part of the numerical value in parentheses.
By adding 0.5 to the numerical value corresponding to the expression and extracting the integer part, an integer value obtained by rounding the numerical value corresponding to Expression 9 is obtained.

Integer value M 'obtained as a result of the operation of equation (10)
Is the number of sampling points per n basic periods of a musical tone having the above-mentioned designated pitch when the musical tone is sampled at the sampling period T.

Subsequently, the analog tone signal having an arbitrary pitch is A / D converted at the sampling period T to obtain PCM tone waveform data, as in the case of the first operation principle described above.

Further, the user gives an estimate of a waveform section for n basic cycles starting from the zero-cross position and ending at the zero-cross position in the continuous sound section of the PCM musical tone waveform data obtained by the A / D conversion, and Specify the start and end of the section.

Subsequently, the loop start initial point and the loop end initial point are calculated in accordance with the above specification, and the loop section is extracted on the basis thereof and the loop section length is calculated. The operation principle is the same as that of the above.

In the sampling period T 'calculated by the following equation (11), the loop section for the above-mentioned n basic periods is set so that the leading zero-cross position and the trailing zero-cross position are included as sampling points. Resampled.

[0137]

[Equation 11]

As a result, loop waveform data including M ′ + 1 samples including the sampling points of the zero cross positions at both ends of the loop section for n basic cycles is generated, and the loop start point and the end point thereof are generated. A certain loop end point exactly matches the leading zero-cross position and the trailing zero-cross position of the waveform.

The loop waveform data of n basic periods composed of M '+ 1 samples generated in this way is saved in a memory or the like. Of the saved loop waveform data of M '+ 1 samples, M' samples from the sampling point at the leading zero-cross position of the loop section to the sampling point one sample before the sampling point at the zero-cross position at the end of the loop section. Is reproduced in a loop at the sampling period T, so that a musical tone substantially matching the designated pitch within a range that does not include loop noise and has no auditory problem. can get. <Specific operation (part 2) of embodiment of the present invention> A specific operation 2 of the embodiment of the present invention for generating loop waveform data based on the above-described second operation principle of the present invention will be described.

First, the user divides the reference pitch of A4 and one basic period corresponding to the reference pitch, for example, T = 1 /
In addition to determining the sampling period T of 5304 [seconds], an arbitrary pitch is designated.

Next, the user determines an integer n such that the value of equation 9 becomes an integer within a predetermined allowable error range. FIG. 22 shows that when the reference pitch is A4 = 442 [Hz] and the sampling period T that divides one basic period corresponding to the reference pitch is 1/5304 [seconds], an accurate pitch of an arbitrary pitch is obtained. A frequency, an example of an integer n, and a pitch frequency when loop waveform data for the n basic cycles are loop-reproduced at a sampling cycle T are shown. Based on this figure, the user can select an integer n corresponding to the designated pitch.
Is determined and input from the key input unit 201 of FIG. It should be noted that the table corresponding to FIG.
03, and the integer n may be automatically calculated according to the pitch specified by the user.

Next, the control / processing unit 205 shown in FIG. 2 performs the following based on the designated pitch, the integer n, and the sampling period T.
Based on equation (10), the number of sampling points M 'per n basic periods of the musical tone when the musical tone having the designated pitch is sampled at the sampling period T is calculated. This sampling point number M 'is stored in the RAM 204 of FIG.

Subsequently, analog tone signals having an arbitrary pitch are A / D-converted at a sampling period T based on the user's operation on the key input unit 201, as in the case of the first specific operation described above. , PCM musical tone waveform data is stored in a hard disk or floppy disk installed in the waveform input / output unit 206 of FIG.
01 is stored.

Thereafter, as in the case of the first specific operation described above, the control / processing unit 205 generates a loop waveform shown by the operation flowchart of FIG. 3 based on the user's operation on the key input unit 201. The control processing program to start.

The processing in step S301 in FIG. 3 is the same as in the first specific operation described above. At the timing when step S302 in FIG. 3 is executed, the user sets the P displayed on the display 102 of the display unit 202.
In the continuous sound section of the CM musical sound waveform data, a waveform section for n basic cycles starting from the zero-cross position and ending at the zero-cross position is registered, and the start and end of the waveform section are designated.

The processing of steps S303 to S308 in FIG. 3 is exactly the same as the specific operation 1 described above. In the next step S309 in FIG. 3, resampling including the loop start point LST and the loop end point LET as sampling points is performed by the following equation corresponding to equation (11) of the <second operation principle of the present invention>. A sampling period T 'for performing the calculation is calculated.

[0147]

(Equation 12)

Here, M 'is a value of a tone having an arbitrary pitch (for example, G4) when the tone is sampled at the sampling period T, as described in the second principle of the present invention. This is the number of sampling points per n basic periods. As described above, the number of sampling points is calculated based on Expression 10 at the start of the loop waveform data generation processing.
It is stored in advance in M204.

The processing in steps S310 to S312 in FIG. 3 is exactly the same as that in the first specific operation described above. As a result, the sampling points at the zero cross positions at both ends of the loop section for n basic cycles are included and M ′ + 1
Waveform data including loop waveform data composed of samples is generated, and a loop start point LSP which is a start point of a loop section and a loop end point LEP which is an end point thereof are, for example, as shown in FIG. Exactly matches the zero-crossing position of the.

The waveform data including the loop waveform data of n basic cycles composed of M '+ 1 samples generated in this way is later stored in a ROM of an electronic musical instrument, and is used as a tone sound source. In this case, the waveform data before the loop section is read out from the memory as it is at the sampling period T and reproduced, so that, for example, after the attack sound section is reproduced, the loop start point LSP
Is loop-reproduced in the sampling period T from n 'basic cycles consisting of M' samples from the sampling point one sample before the loop end point LEP to a loop-free loop noise, and an auditory problem. Within this range, it is possible to produce a musical tone that substantially matches the designated pitch.

In the second specific operation of the embodiment of the present invention, as can be seen from FIG. 22, the designated pitches are, for example, G3, A3, D4, E4, A4 (reference pitch), D
In the case of 5, E5, A5, etc., since n = 1, the equations (10), (11) and (12) are based on the above-mentioned equations (1), (2) and (6), respectively. This is equivalent to replacing the pitch with the specified pitch. Therefore, in this case,
Except for the addition of the pitch designation operation, substantially the same operation as the first specific operation of the embodiment of the present invention described above is performed.

[0152]

According to the present invention, the loop start point and the loop end point can be respectively adjusted to the zero-cross position by a simple operation in the process of creating a loop waveform for generating a continuous sound, and the loop noise is reduced. , And within a range in which there is no problem in hearing, a waveform substantially matching the designated arbitrary pitch can be obtained.

[Brief description of the drawings]

FIG. 1 is an external view of one embodiment of the present invention.

FIG. 2 is an overall configuration diagram of one embodiment of the present invention.

FIG. 3 is an operation flowchart of a loop waveform generation process.

FIG. 4 is a waveform diagram showing a relationship between a waveform before and after a waveform shift and a zero-cross position.

FIG. 5 is a principle diagram of a method for calculating a time from an initial point of a loop start to a zero-cross position immediately after the initial point (part 1);
It is.

FIG. 6 is a principle diagram of a method for calculating a time from a loop start initial point to a zero cross position immediately after the initial point (part 2);
It is.

FIG. 7 is a principle diagram of a method for calculating a time from a loop start initial point to a zero cross position immediately after the loop start point (part 3);
It is.

FIG. 8 is a principle diagram of a calculation method of a time from an initial point of a loop start to a zero cross position immediately after the loop start point (part 4);
It is.

FIG. 9 is a principle diagram of a method for calculating a time from a loop start initial point to a zero cross position immediately after the initial point (part 5).
It is.

FIG. 10 is a detailed operation flowchart of step S304 in FIG. 3;

FIG. 11 is an explanatory diagram of linear interpolation.

FIG. 12 is an explanatory diagram of Lagrange interpolation using three points.

FIG. 13 is a diagram (No. 1) showing values substituted into Expression 3;

FIG. 14 is a detailed operation flowchart relating to a waveform shift process in step S305 in FIG. 3;

FIG. 15 is a diagram (No. 2) illustrating values substituted for Expression 3.

FIG. 16 is a waveform diagram showing a relationship between waveforms before and after resampling and a zero-cross position.

FIG. 17 is a detailed operation flowchart relating to a resampling process in step S310 of FIG. 3;

FIG. 18 is an operation timing chart showing timings of sample values before and after resampling.

FIG. 19 is a diagram showing an interpolation value f (i) and sample values of shifted waveform data before and after the interpolation value f (i).

FIG. 20 is a diagram (No. 3) illustrating values substituted for Expression 3.

FIG. 21 is a waveform diagram when a sound other than A4 is sampled at a sampling period T.

FIG. 22 is a diagram showing an accurate pitch frequency for each pitch, a value of an integer n, and a pitch frequency during loop reproduction.

FIG. 23 is an explanatory diagram of a specific operation (part 2) of the embodiment of the present invention.

FIG. 24 is an explanatory diagram of loop noise in a conventional example.

[Explanation of symbols]

 101 source waveform file 102 display 103 personal computer 104 keyboard 105 mouse 106 created waveform file 201 key input unit 202 display unit 203 ROM 204 RAM 205 control / processing unit 206 waveform input / output unit

Claims (6)

(57) [Claims] 1. A loop waveform for generating digital waveform data for one repetitive waveform unit to be repeatedly reproduced in the first sampling period from digital waveform data sampled in a first sampling period as loop waveform data. In the generation device, a pitch designation means for designating a pitch at the time of the repetitive reproduction, and a first cycle which is an integral multiple of one basic cycle corresponding to the pitch designated by the pitch designation means, Regarding a cycle that is divisible within a predetermined allowable error range by a sampling cycle, a sampling point number calculation that calculates a number obtained by dividing the predetermined cycle within the predetermined allowable error range by the first sampling cycle as a sampling point number. Means, based on the digital waveform data, a waveform section for the predetermined period. Loop section extracting means for extracting, as a loop section, a section from the first zero-cross position, which is the beginning of the waveform section, to the second zero-cross position, which is the end of the waveform section for the predetermined period; The data is stored in the first
By re-sampling at a second sampling period obtained by dividing the time length of the loop section by the number of the sampling points so that the zero-cross position and the second zero-cross position are included as sampling points, Resampling means for generating loop waveform data; and a loop waveform generation device. 2. A loop waveform for generating, as loop waveform data, digital waveform data for one repetitive waveform unit to be repeatedly reproduced at the first sampling period, from digital waveform data sampled at a first sampling period. In the generation method, a pitch at the time of the repetitive reproduction is specified, and a predetermined cycle of an integral multiple of one basic cycle corresponding to the specified pitch within a predetermined allowable error range by the first sampling cycle. Regarding the divisible period, a number obtained by dividing the predetermined period within the predetermined allowable error range by the first sampling period is calculated as the number of sampling points, and based on the digital waveform data, From the first zero-cross position, which is the beginning of the waveform section, A section up to a second zero-cross position at the end is extracted as a loop section, and the digital waveform data in the loop section is extracted as the first section.
By re-sampling at a second sampling period obtained by dividing the time length of the loop section by the number of the sampling points so that the zero-cross position and the second zero-cross position are included as sampling points, A method for generating loop waveform data, the method comprising: generating loop waveform data. 3. A re-sampling process of said digital waveform data by said re-sampling means in said second sampling cycle is performed by converting a sample value for each time position advanced in said second sampling cycle into said first sampling cycle. The loop waveform generation device according to claim 1, wherein the calculation is performed by interpolation using a sampling point near the time position among the sampling points of the digital waveform data sampled in ( 1 ).
Place . 4. The loop waveform generation device according to claim 3, wherein the interpolation calculation is an interpolation calculation using a Lagrange's interpolation function.
Place . 5. The digital waveform data according to claim 5, wherein
The resampling process with the sampling period of 2
The sample at each time position to advance in the second sampling cycle
Values are sampled at said first sampling period
The sampling point of the digital waveform data
By using an interpolation operation using the one near the time position,
The process of calculating 3. The loop waveform generation method according to claim 2, wherein:
Law. 6. The interpolation operation according to claim 1, wherein the Lagrange interpolation function is used.
Is an interpolation operation using numbers. 6. The method according to claim 5, wherein the loop waveform is generated.
Law.