JP3714397B2 - Waveform data processing apparatus and waveform data processing method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a waveform data processing apparatus and method for correcting waveform data composed of a plurality of waveform sample values stored in a waveform memory and used to generate a musical tone signal.
[0002]
[Prior art]
Conventionally, one musical sound waveform is divided into a plurality of temporally different parts, for example, divided into an attack part and a sustain part, and waveform data consisting of a plurality of waveform sample values representing the musical sound waveform of each part is independent. A musical tone signal generating apparatus that stores data in a waveform memory and continuously reads out the waveform data of each part to generate one musical tone waveform signal is well known. In this case, the pitch of the musical sound signal generated based on the waveform data of each part read at the same rate may be different, and in order to avoid this, when reading the waveform data of each part, the reading is performed. Musical tone signal generators that allow the rate to be adjusted for each part have also appeared.
[0003]
[Problems to be solved by the invention]
However, even in a musical tone signal generator that can adjust the readout rate for each part, the change of the readout rate is accurately synchronized with the switching of the readout of the waveform data of each part and the readout rate is accurately changed. It is difficult to do this, and there is a problem that the pitch of the tone signal generated when switching the reading of the waveform data of each part becomes discontinuous. Further, in a musical tone signal generating apparatus that does not have the function of adjusting the reading rate for each part, there is a possibility that a large pitch fluctuation occurs when switching the reading of the waveform data of each part.
[0004]
[Outline of the present invention]
The present invention has been made in order to cope with the above-described problem. The object of the present invention is to correct the waveform data of each part stored in the waveform memory in advance, and to perform the waveform data of each part at the same rate. Provided is a waveform data processing apparatus and method for preventing pitch fluctuations even when a musical tone signal is generated by continuously reading.
[0005]
In order to achieve the above object, the structural features of the present invention are: Music waveform data that has been sampled at a predetermined sampling rate and has an attack part and a sustain part is input and read sequentially after the attack waveform data and the attack waveform data are read out to form the attack part of the tone signal. And loop waveform data that forms the sustain part of the musical sound signal Waveform data processor In the above-mentioned input musical sound waveform data, musical sound waveform data from the attack start point specified by the user to the attack end point is stored as attack waveform data, and from the loop start point specified by the user to the loop end point Waveform memory for storing the musical tone waveform data up to the above as loop waveform data, and an attack unit for calculating the pitch of the attack portion of the musical tone waveform signal obtained by reproducing the attack waveform data stored in the waveform memory by frequency analysis The pitch calculation means, the pitch of the loop portion of the musical sound waveform signal reproduced by repeatedly reading out the loop waveform data stored in the waveform memory, the pitch of the loop portion obtained by frequency analysis, the loop start point Loop portion pitch calculating means for calculating using a loop size representing the number of samples from the loop end point and the predetermined sampling rate, and at least one of the first attack waveform data and the loop waveform data, The pitch of the attack part and the pitch of the loop part of the calculated tone waveform signal Change the sampling rate according to the ratio of Resampling And resampling means. Another structural feature of the present invention is also a method for realizing the function of each of the above means.
[0006]
In the present invention configured as described above, Sampling rate changed according to the ratio of the pitch of the attack portion of the musical sound waveform signal obtained by reproducing the attack waveform data and the pitch of the loop portion of the musical sound waveform signal reproduced by repeatedly reading the loop waveform data At least one of the attack waveform data and the loop waveform data is resampled. It will be stored in the waveform memory. Therefore, when generating a musical sound signal, Attack waveform data as well as loop Even if waveform data is read continuously at the same rate, attack From waveform data loop Pitch fluctuation when switching waveform data can be avoided, Attack waveform data as well as loop When generating a musical sound signal using waveform data, it is possible to always provide a musical sound signal in which pitch fluctuations associated with the switching are suppressed as much as possible.
[0007]
DETAILED DESCRIPTION OF THE INVENTION
DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 schematically shows a musical sound waveform processing apparatus according to the embodiment in a block diagram.
[0008]
This musical tone waveform processing apparatus includes a CPU 11, a timer 12, a ROM 13, and a RAM 14 that are connected to a bus 10 and constitute a computer main body. The CPU 11 controls the entire musical sound waveform processing apparatus and executes various programs. The timer 12 provides the CPU 11 with information related to the time necessary for executing the program. The ROM 13 stores a part of the program and various information necessary for generating musical sounds. The RAM 14 temporarily stores a part of the program and information necessary for executing the program.
[0009]
The musical tone waveform processing apparatus also includes a waveform memory 21, an access management circuit 22, a writing circuit 23, a buffer circuit 24, a tone generator circuit 25, and a sound system 26. The writing circuit 23, the buffer circuit 24, and the sound source circuit 25 are connected to the bus 10. The waveform memory 21 is composed of a rewritable nonvolatile memory (memory that can store data even when power supply is cut off), and stores waveform data representing a musical sound waveform and accompanying data accompanying the waveform data. is there. The access management circuit 22 manages writing and reading of waveform data and accompanying data to the waveform memory 21.
[0010]
The writing circuit 23 includes an A / D converter, samples an analog external sound signal supplied to the external sound input terminal 27 at a predetermined sampling frequency (standard sampling frequency) fs, and outputs the sampled external sound signal. The sound signal is A / D converted and supplied to the access management circuit 22. The external sound signal may be input from a microphone or recorded on a tape recorder. The writing circuit 23 can also input an external sound signal that has been converted into a digital signal in advance, and in this case, the A / D conversion is not necessary. The buffer circuit 24 temporarily stores waveform data between the bus 10 and the access management circuit 22.
[0011]
The tone generator circuit 25 generates and outputs a musical sound signal according to an instruction from the CPU 11. In response to the instruction, the tone generator circuit 25 reads out the waveform data stored in the waveform memory 21 via the access management circuit 22 and reads out the same. A musical tone signal is formed and output based on the waveform data. In this case, the waveform data is composed of two waveform data of the attack part and the loop part for one musical tone, the waveform data of the attack part is read sequentially from the head, and then the waveform data of the loop part is repeatedly read. Then, the read waveform data is D / A converted and output as a tone signal. The sound system 26 includes an amplifier, a speaker, and the like, and converts the musical sound signal output from the sound source circuit 25 into an acoustic signal and outputs the sound signal.
[0012]
The musical tone waveform processing apparatus also includes a display 31, a panel switch 32, an interface 33, and a drive circuit 34 connected to the bus 10. The display 31 is controlled by the CPU 11 to display instructions for the user, graphics representing various data, characters, and the like. The panel switch 32 is used for inputting a user instruction and various data to the CPU 11. The interface 33 is connected to other musical tone generators, electronic musical instruments, sequencers, automatic performance devices, personal computers, communication lines, etc., and is used to transmit and receive various information including MIDI information. The drive circuit 34 can write and read various data and programs to various disks such as a hard disk, a compact disk, and a flexible disk as an external memory.
[0013]
In this musical sound waveform processing apparatus, the program shown in FIG. 3 and a program (not shown) which will be described later may be stored in the ROM 13 in advance, and what is recorded on the disk 35 is stored in the RAM 14. It may be used after writing. Alternatively, the program may be used after being input from the outside via the interface 33 and written to the RAM 14 or the disk 35.
[0014]
Next, the operation of the musical tone waveform processing apparatus configured as described above will be described. First, as shown in step 102 of FIG. 2, the user operates the panel switch 32 in order to acquire musical sound waveform data for one sound from the outside. In this case, an external sound signal having a desired pitch frequency (for example, a natural musical instrument sound signal) is input to the writing circuit 23 via the external sound input terminal 27 by playing a natural musical instrument or playing a tape recorder. Further, an external sound signal (external sound waveform data) for one sound recorded on the disk 35 as an external recording medium may be read by the drive circuit 34 and supplied to the writing circuit 23 via the bus 10. .
[0015]
If the input external sound signal is in an analog format, the writing circuit 23 samples the external sound signal input at a predetermined sampling frequency (standard sampling frequency) fs and performs A / D conversion. The converted external sound signal is written into the waveform memory 21. If the input external sound signal is in digital format, the external sound signal is written in the waveform memory 21 as it is. In this case, there is no problem as long as the external sound signal converted into the digital format is sampled at the same sampling frequency as the sampling frequency fs. Data representing the sampling frequency is written in the waveform memory 21 together with the sound signal. The external sound signal may be stored in another writable memory such as the buffer circuit 24 and the RAM 14 instead of the waveform memory 21.
[0016]
Next, in step 104, the user inputs a note number NN corresponding to the pitch frequency of the input external sound signal using the panel switch 32, and displays a waveform diagram of the input external sound signal on the display. 31 is displayed. When the note number NN is input together with the external sound signal, the input of the note number NN using the panel switch 32 is omitted. Next, in step 104, the user performs the cut-out operation of the attack portion and the loop portion using the panel switch 32 while looking at the waveform diagram of the external sound signal on the display device 31 in units of each sample value.
[0017]
In this cut-out operation, the user designates the attack start point P1, attack end point P2, loop start point P3, and loop end point P4 of the external sound signal on the display 31. The attack start point P1 is designated as a timing position at which the amplitude envelope of the musical sound signal starts rising from “0”. The attack end point P2 is designated as a timing position (sustain portion start timing position) at which the change of the amplitude envelope almost disappeared and began to stabilize. The loop start point P3 and the loop end point P4 are designated as the start and end timing positions of the section where the tone signal waveform repeatedly changes in the sustain portion. In this case, one period or a plurality of periods of the tone signal waveform of the sustain section may be designated as the section.
[0018]
By this designation, the CPU 11 executes a program (not shown), among a series of sample values representing an external sound signal written in the waveform memory 21 (or other writable memory such as the buffer circuit 24 and the RAM 14). The sample values belonging to the designated attack start point P1 to attack end point P2 are written in the waveform memory 21 as attack waveform data ADT, and the sample values belonging to the loop start point P3 to loop end point P4 are written to the loop waveform data. Write to the waveform memory 21 as LDT. At this time, the start and end addresses of the attack waveform data ADT, the start and end addresses of the loop waveform data LDT, and the inputted note number NN are attached to the attack waveform data ADT and the loop waveform data LDT to be a waveform memory. Write to 21. Hereinafter, the start and end addresses of these attack waveform data ADT are referred to as an attack start address and an attack end address, and the start and end addresses of the loop waveform data LDT are referred to as a loop start address and a loop end address. Note that a series of sample values representing the external sound signal are deleted at this point, but when the sampling frequency of the external sound signal is written together with the external sound signal as described above, the sampling frequency is set to the sampling frequency. The data is written in the waveform memory 21 together with each data.
[0019]
Here, the necessity of adjusting the pitch of the attack part and the loop part (sustain part) by resampling described later will be described. In the formation of the musical tone signal, the attack waveform data ADT is sequentially read out, and then the loop waveform data LDT is repeatedly read out to form the musical tone signal. However, in this case, if both the read rates of the attack waveform data ADT and the loop waveform data LDT are the same, a large pitch fluctuation may occur at the time of switching from the attack portion of the generated musical sound signal to the sustain portion (loop portion). is there.
[0020]
That is, in the attack portion of the tone signal that is formed, a tone signal having a frequency that is exactly proportional to the ratio between the sampling rate of the external tone signal and the readout rate is obtained. However, since the waveform data is cut out in units of sample values, in the sustain portion (loop portion), as shown in FIG. 4B, the loop start point P3 and the loop end point P4 are the sustain of the external sound signal. It does not exactly match the front and rear edges of the repetitive waveform of the part. For example, when the loop start point is set to P3, the ideal loop end point corresponding to the point P3 is P4 *, but the sampling frequency fs is not exactly an integral multiple of the external sound signal. Waveform data from P3 to loop end point P4 * cannot be cut out, and an error Δt shown in FIG. 4B occurs. This is apparent from the fact that the newly formed musical sound signal does not have the same frequency as the external sound signal even when the loop waveform data LDT is read out with the read rate being the same as the sampling rate of the external sound signal. Therefore, if the attack waveform data ADT is read out at a constant readout rate (for example, the same readout rate as the sampling rate of the external sound signal) and then the loop waveform data LDT is repeatedly read out, the sustain from the attack portion of the generated musical sound signal is sustained. There is a possibility that a large pitch fluctuation may occur at the time of switching to the portion (loop portion).
[0021]
After the processing of step 104 in FIG. 2, the user instructs the resampling of the waveform data stored in the waveform memory 21 using the panel switch 32 in step 106. In response to this instruction, the CPU 11 starts the resampling program of FIG. 3 at step 200, and determines whether or not the loop setting processing is performed on the external sound signal written to the waveform memory 21 at step 202. That is, it is determined whether or not the attack waveform data ADT and the loop waveform data LDT are stored in the waveform memory 21. This is because the above-mentioned loop waveform data setting processing is applied to musical tone waveforms that are damped and whose pitch is not clear, that is, tones that are difficult to specify a loop (for example, waveforms related to percussion instrument sounds such as bass drums and sound effects such as gunshots). This is because the external sound signal is stored as it is. If the loop setting process has not been performed, “NO” is determined in step 202, and the execution of the resampling program is terminated in step 204.
[0022]
On the other hand, if the loop setting process has been performed on the external sound signal written in the waveform memory 21, “YES” is determined in the step 202, and the program proceeds to the steps 206 to 210. In step 206, sample values for a plurality of cycles in a portion close to the attack end are extracted from a series of sample values constituting the attack waveform data ADT, and the waveform data represented by the extracted sample values is used as the attack end waveform. The data is AEDT. Specifically, a plurality of sample values that are traced back by a plurality of cycles from the attack end address stored in the waveform memory 21 are read, and the waveform data represented by the read sample values is used as the attack end waveform data AEDT. The attack end waveform data AEDT is subjected to a fast Fourier transform (FFT) process, and the waveform signal represented by the waveform data AEDT is subjected to frequency analysis to detect a frequency component included in the waveform signal. Thereby, in the normal case, an analysis result having peaks corresponding to a plurality of specific frequencies as shown in FIG. 5 is obtained. The peak may not be detected due to many noise components.
[0023]
Next, at step 208, the note number NN stored in the waveform memory 21 along with the attack and loop waveform data ADT, LDT is read out, and the pitch Fnn (corresponding to the note number NN is read from the frequency analysis result. A peak in a frequency band in the vicinity of the pitch frequency of the external sound signal) is detected. If the corresponding peak is detected, the detected peak is set as the fundamental wave peak PK1, the frequency at which the peak PK1 appears is set as the fundamental wave frequency PF1, and the level of the peak PK1 is set as the fundamental wave level PL1. . The pitch Fnn corresponding to the note number NN is stored in advance in the ROM 13 or the waveform memory 21 corresponding to each note number NN.
[0024]
Next, in step 210, a peak in a frequency band near twice the pitch Fnn corresponding to the note number NN is detected from the frequency analysis result. If a corresponding peak is detected, the detected peak is set as the second harmonic peak PK2, the frequency at which the peak PK2 appears is set as the second harmonic frequency PF2, and the level of the peak PK2 is set as the second harmonic level PL2. .
[0025]
After the process of step 210, in step 212, the next determination process based on the detection results of steps 206 to 210 is performed. First, when neither the fundamental wave nor the second harmonic peak PK1, PK2 is detected, the program proceeds to step 214 by the determination process of step 212. In step 214, the attack pitch AP is set to the pitch Fnn (pitch frequency of the external sound signal) corresponding to the note number NN derived in the same manner as described above, and the harmonic number HN is set to “1”.
[0026]
If the second overtone peak PK2 is not detected and only the fundamental wave peak PK1 is detected by the processing of steps 206 to 210, the program proceeds to step 216 by the determination processing of step 212. In step 216, the attack pitch AP is set to the fundamental frequency PF1, and the harmonic number HN is set to “1”. If the fundamental wave peak PK1 is not detected by the processing of steps 206 to 210 and only the second harmonic peak PK2 is detected, the program proceeds to step 218 by the determination processing of step 212. In step 218, the attack pitch AP is set to the second harmonic frequency PF2, and the harmonic number HN is set to “2”.
[0027]
If both the fundamental wave and the second harmonic peaks PK1 and PK2 are detected by the processing of steps 206 to 210, the program proceeds to step 220 by the determination processing of step 212. In step 220, it is determined which of the peaks is dominant. In this case, the pitch stability and continuity determined based on the magnitude relationship between the fundamental wave level PL1 and the second harmonic level PL2, the level around the fundamental wave and the second harmonic peaks PK1, PK2, and the like are determined. The higher the effectiveness of, the more effective.
[0028]
If it is determined that the fundamental wave peak PK1 is influential, the program proceeds to step 220 under the determination of “YES” in step 220. In step 220, as in step 216, the attack pitch AP is set to the fundamental frequency PF1, and the harmonic number HN is set to “1”. If it is determined that the second overtone peak PK2 is dominant, the program proceeds to step 222 under the determination of “NO” in step 220. In step 222, as in the case of step 218, the attack pitch AP is set to the second harmonic frequency PF2, and the harmonic number HN is set to “2”.
[0029]
By the processing of these steps 206 to 222, the pitch (frequency) of the attack part (more precisely, the attack end part) of the external sound signal is detected. In other words, when the attack waveform data ADT stored in the waveform memory 21 is read out at the same reading rate as the sampling frequency fs of the external sound signal to form a musical sound signal, the attack portion of the same musical sound signal (more precisely, the attack is performed). The pitch (frequency) of the end portion is detected.
[0030]
Next, the pitch (frequency) of the musical tone signal when the musical tone signal is formed by reading the loop waveform data LDT at the same readout rate is detected by the processing of steps 224 to 228.
[0031]
In step 224, the pitch Fnn corresponding to the note number NN read from the waveform memory 21 is derived in the same manner as in steps 206 and 208, and the sampling frequency fs obtained by sampling the external sound signal is calculated using the derived pitch Fnn. By dividing, the number of samples fs / Fnn per cycle of the external sound signal is calculated. Next, in step 224, the loop start address and loop end address of the loop waveform data LDT stored in the waveform memory 21 are read out, and the loop size LS (the set loop waveform data LDT of the set loop waveform data LDT is read from both of the read addresses). Calculate the number of samples). Then, the wave size WN (= LS · Fnn / fs) included in the loop waveform data LDT is calculated by dividing the loop size LS by the number of samples fs / Fnn in one period. Specifically, the first decimal place of the calculation result (LS · Fnn / fs) may be rounded off to calculate an integer value.
[0032]
In step 226, the loop frequency LP (= fs / LS) indicating the pitch (frequency) per loop of the loop portion is calculated by dividing the sampling frequency fs by the calculated loop size LS. In step 228, the pitch (frequency) LWP (= WN · LP) of the waveform signal included in the loop portion is calculated by multiplying the calculated wave number WN by the loop pitch LP. As a result, when the tone signal is formed by repeatedly reading out the loop waveform data LDT stored in the waveform memory 21 at the same reading rate as the sampling frequency fs of the external sound signal, the pitch of the loop portion (sustain portion) of the same tone signal is formed. (Frequency) is detected.
[0033]
After calculating the pitch (frequency) AP, LWP of the musical sound signal based on the both waveform data ADT, LDT in the case of reading at the same reading rate in this way, the CPU 11 determines in step 230 that the harmonic number HN is “1”. It is determined whether or not.
[0034]
If the harmonic number HN is “1”, that is, the pitch (frequency) of the fundamental wave is detected as the pitch (frequency) of the attack portion, “YES” is determined in step 230, and the program proceeds to step 232. In step 232, the attack waveform data ADT stored in the waveform memory 21 is resampled according to the pitch AP of the attack portion and the pitch LWP of the loop portion. That is, the attack waveform data ADT stored in the waveform memory 21 and sampled at the sampling frequency fs is resampled at the resampling frequency fs · AP / LWP corresponding to the ratio between the two pitches AP and LWP (resampling). Then, the attack waveform data ADT previously stored in the waveform memory 21 is rewritten to the resampled attack waveform data ADT ′. When the resampled attack waveform data ADT ′ is written to the waveform memory 21, the data ADT ′ is written forward from the attack end address of the waveform memory 21. In step 236, the attack start address in the waveform memory 21 is rewritten to the head address of the attack waveform data ADT '.
[0035]
On the other hand, if the harmonic number HN is “2”, that is, if the pitch (frequency) of the second harmonic is detected as the pitch (frequency) of the attack portion, “NO” is determined in step 230 and the program proceeds to step 234. . In step 234, the attack waveform data ADT stored in the waveform memory 21 is resampled at a resampling frequency fs · AP / 2 · LWP corresponding to “½” of the resampling frequency fs · AP / LWP. . This is because the pitch corresponding to the second overtone of the external sound signal is detected in the pitch detection of the attack portion. Then, the attack waveform data ADT previously stored in the waveform memory 21 is rewritten to the attack waveform data ADT ′ resampled as described above. Also in this case, in step 236, the attack start address in the waveform memory 21 is rewritten to the head address of the attack waveform data ADT ′ in the same manner as described above.
[0036]
By such resampling processing, the attack waveform data ADT (FIG. 6A) stored in the waveform memory 21 is rewritten to the resampled attack waveform data ADT ′ as shown in FIG. 6B. . The example of FIG. 6B shows an example in which the pitch of the attack part (accurately, the attack end part) is higher than the pitch of the loop part. In this case, the pitch AP of the attack portion is larger than the pitch LWP of the loop portion, and the re-sampling frequency fs · AP / LWP is larger than the original sampling frequency fs. Therefore, the sample of the attack waveform data ADT ′ after the resampling The number is larger than the original attack waveform data ADT by an amount corresponding to the ratio AP / LWP of both pitches. Conversely, if the pitch of the attack portion is lower than the pitch of the loop portion, the number of samples of the attack waveform data ADT ′ after resampling is an amount corresponding to the ratio AP / LWP of both pitches than the original attack waveform data ADT. Less.
[0037]
In the above description, the standard sampling frequency fs determined in advance is used. However, when the sampling frequency is written in the waveform memory 21 when sampling the external sound signal as described above, the same is true. The written sampling frequency may be used in place of the sampling frequency fs.
[0038]
Then, after the processing in step 236, execution of the resampling program is terminated in step 238. When a new external sound signal is processed as a musical sound waveform, the attack waveform data ADT, the loop waveform data LDT, and data associated therewith are stored in the waveform memory 21 in accordance with the operation shown in FIG. Remember me.
[0039]
When the waveform data written in the waveform memory 21 is read in this way to form a musical tone signal, the CPU 11 instructs the tone generator circuit 25 to generate a musical tone having a desired pitch frequency. In response to this instruction, the sound source circuit 25 performs re-sampling from the waveform memory 21 via the access management circuit 22 at a read rate corresponding to the pitch frequency (for example, the same read rate as the original sampling frequency fs). The attack waveform data ADT ′ is read out, and then the loop waveform data LDT is repeatedly read out. A musical tone signal is formed on the basis of the read attack waveform data ADT ′ and loop waveform data LDT, and is output via the sound system 26. In this case, when the release start is instructed by the instruction of the CPU 11, the sound source circuit 25 may gradually attenuate and output the amplitude envelope of the sample value represented by the loop waveform data LDT.
[0040]
As a result of the musical sound signal being generated in this way, when the pitch of the attack part (accurately, the attack end part) is higher than the pitch of the loop part (sustain part), it is stored in the waveform memory 21 as described above. Since the number of samples of the resampled attack waveform data ADT ′ is larger in proportion to the pitch ratio AP / LWP than the number of samples of the original attack waveform data ADT, the attack waveform data ADT ′ and Even if the loop waveform data LDT is continuously read, it is possible to avoid the pitch fluctuation at the switching timing between the attack part and the loop part. On the contrary, when the pitch of the attack portion (accurately, the attack end portion) is lower than the pitch of the loop portion (sustain portion), the number of samples of the resampled attack waveform data ADT ′ is the number of samples of the original attack waveform data ADT. Since the number is smaller in proportion to the pitch ratio AP / LWP than the number, even if the attack waveform data ADT ′ and the loop waveform data LDT are continuously read out at the same readout rate, the switching timing between the attack part and the loop part It is possible to avoid pitch fluctuations in
[0041]
As can be understood from the above operation description, according to the above embodiment, the attack waveform data ADT and the loop waveform data LDT formed by sampling the external sound signal are processed by the processing of steps 206 to 222 and the processing of steps 224 to 228. The pitches (frequencies) AP and LWP of each musical tone signal when read out and formed at the same rate were detected. Then, by the processing of steps 230 to 236, the attack waveform data ADT 'is resampled according to the ratio AP / LWP of both pitches AP and LWP, thereby forming new attack waveform data ADT' in which the number of samples is changed. The original attack waveform data ADT was rewritten with the resampled attack waveform data ADT ′. As a result, even if the attack waveform data ADT ′ and the loop waveform data LDT are read at the same rate and continuously read at the rate, no pitch difference is generated in the tone signal based on both waveform data ADT ′ and LDT. Pitch fluctuation at the time of switching between the attack part and the loop part can be avoided without taking a method such as switching the reading rate according to the data.
[0042]
In addition, since the pitch of the musical sound signal is detected using a plurality of peaks obtained by FFT processing as in the processing of steps 206 to 222 above, the pitch of the musical sound signal is calculated even when only one of the peaks is detected. it can. If a plurality of peaks can be detected, a more appropriate value can be calculated as the pitch of the musical sound signal using the plurality of detected peaks. As a result, the pitch of the musical sound signal can be detected widely and the pitch can be detected with high accuracy.
[0043]
In the above embodiment, the resampling process is performed on the attack waveform data ADT. However, the resampling process may be performed on the loop waveform data LDT. In this case, both pitches AP and LWP are reversed in steps 232 and 234 of the above embodiment, and the loop waveform data LDT is resampled by fs · LWP / AP and fs · 2 · LWP / AP, respectively. That's fine. FIG. 6C shows such an example. When the pitch of the attack portion (accurately, the attack end portion) is lower than the pitch of the loop portion (sustain portion), a sample of the resampled loop waveform data LDT ′ is shown. The number is smaller in proportion to the pitch ratio LWP / AP than the number of samples of the original loop waveform data LDT. Conversely, when the pitch of the attack portion is higher than the pitch of the loop portion, the number of samples of the resampled loop waveform data LDT ′ is proportional to the pitch ratio LWP / AP rather than the number of samples of the original loop waveform data LDT. And increase.
[0044]
In the above-described embodiment, the waveform data is subjected to FFT processing in steps 206 to 222 so as to detect only the fundamental wave peak PK1 and the second harmonic peak PK2. However, the third harmonic, the fourth harmonic,. The peaks PK3, PK4, etc. relating to many overtones may be detected by FFT processing, and the pitch of the musical sound signal may be detected by selectively and effectively using these detected many peaks. . According to this, even if the fundamental wave peak PK1 and the second harmonic peak PK2 are not detected, the pitch of the musical sound signal can be detected based on the peaks PK3, PK4,. . In addition, by using more peaks in combination, it is possible to improve the pitch detection accuracy of the musical sound signal.
[0045]
Further, in the above embodiment, the pitch (frequency) of the attack part is calculated by the process including steps 206 to 222 using the fast Fourier transform (FFT) process, and the pitch (frequency) of the loop part is calculated as the sampling frequency fs and the loop size. The calculation was performed by the processing including steps 224 to 228 using LS, wave number WN, and the like. However, when connecting waveform data set to loop readout, the pitch for both waveform data may be calculated by the latter process for both waveform data. When connecting waveform data for which loop reading is not set, the pitch for both waveform data may be calculated by the former processing for both waveform data. In this case, instead of performing the FFT processing on the attack end waveform data AEDT of the above embodiment, waveform data corresponding to a predetermined period or a predetermined number of samples are respectively extracted from the connecting points for each waveform data, and the extracted waveform data is obtained. Each may be subjected to FFT processing. In addition, when trying to connect waveform data that is not loop-read after waveform data that is set to loop readout, the pitch for the waveform data that is set to loop readout is calculated by the latter processing, The pitch of waveform data that is not performed may be calculated by the former process. Furthermore, the pitch of the attack part, the loop part, and each of the waveform data divided into a plurality of parts described later may be calculated by any method as long as the reproduction pitch is calculated.
[0046]
In particular, when the waveform data for which loop reading is not set is connected, as described above, the pitch for both waveform data by FFT processing may be calculated for both waveform data. That is, as described above, a plurality of peaks such as the fundamental wave peak PK1 and the second overtone peak PK2, or the third overtone peak PK3 and the fourth overtone peak PK4 according to these peaks are detected, and the detected plurality of peaks are determined. It is good to calculate the pitch for both waveform data. According to this, both waveform data will be connected based on the overtone with high possibility of being recognized as a pitch, and appropriate connection of both waveform data is implement | achieved.
[0047]
Moreover, although the said embodiment solves the problem which arises because the extraction of a loop part was performed per sample value, this resampling process of waveform data is applied widely. That is, when a plurality of waveform data that are stored in the waveform memory 21 at a predetermined sampling rate and are each composed of a plurality of waveform sample values are continuously read and a musical tone signal is formed based on the read waveform data. Widely applicable. Although there is no problem in cutting out the waveform data and the pitch (frequency) of the musical sound signal should be the same, multiple waveform data were collected under different conditions, so the pitch (frequency) of the musical sound signal was May shift slightly. For example, when the attack waveform data ADT and the loop waveform data LDT are collected independently from the natural instrument sound under different environments such as different temperatures and slightly changed tuning conditions, the loop waveform data for the waveform data with different touches This is the case when trying to use LDT in common. Even in such a case, if one of the two waveform data continuously read out as described above is resampled according to the pitch ratio as described above, the musical tone signal based on the two waveform data is used. Playback pitch (frequency) can be adjusted.
[0048]
In the above embodiment, the waveform data is divided into the attack part and the loop part (sustain part), but in addition to the attack part and the loop part (sustain part), the release part is also handled as one part. Release waveform data consisting of a plurality of sample values may be prepared. Further, the attack part, the sustain part, and the release part may be further subdivided.
[Brief description of the drawings]
FIG. 1 is an overall block diagram of a musical sound waveform processing apparatus according to an embodiment of the present invention.
FIG. 2 is a flowchart showing a procedure for creating sound source waveform data by a user.
FIG. 3 is a flowchart of a program executed in response to a resampling instruction in FIG.
4A is a waveform diagram showing an external sound signal, and FIG. 4B is an enlarged waveform diagram showing a front end and a rear end of a loop portion of the external sound signal.
FIG. 5 is a spectrum diagram obtained by FFT processing in FIG. 3;
FIG. 6 is an amplitude envelope waveform diagram of a musical sound waveform for explaining resampling.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 ... Bus, 11 ... CPU, 13 ... ROM, 14 ... RAM, 21 ... Waveform memory, 23 ... Write circuit, 25 ... Sound source circuit, 31 ... Display, 32 ... Panel switch.

Claims (2)

Music waveform data that has been sampled at a predetermined sampling rate and has an attack part and a sustain part is input and read sequentially after the attack waveform data and the attack waveform data are read out to form the attack part of the tone signal. In the waveform data processing device for generating the loop waveform data constituting the sustain portion of the musical sound signal ,
Among the input musical sound waveform data, the musical sound waveform data from the attack start point specified by the user to the attack end point is stored as attack waveform data, and from the loop start point specified by the user to the loop end point. A waveform memory for storing musical sound waveform data as loop waveform data;
Attack part pitch calculating means for calculating the pitch of the attack part of the musical sound waveform signal obtained by reproducing the attack waveform data stored in the waveform memory by frequency analysis;
The pitch of the loop portion of the musical sound waveform signal reproduced by repeatedly reading out the loop waveform data stored in the waveform memory, the pitch of the loop portion obtained by frequency analysis, the number of samples from the loop start point to the loop end point Loop part pitch calculating means for calculating using a loop size representing the predetermined sampling rate;
Wherein at least either one of the attack waveform data and the loop waveform data, the calculated resampled by changing the sampling rate according to the ratio of the pitch of the loop portion of the attack portion of the musical tone waveform signal And a resampling means for performing the waveform data processing. Music waveform data that has been sampled at a predetermined sampling rate and has an attack part and a sustain part is input and read sequentially after the attack waveform data and the attack waveform data are read out to form the attack part of the tone signal. In the waveform data processing method for generating the loop waveform data constituting the sustain portion of the musical sound signal ,
Of the input musical sound waveform data, musical sound waveform data from the attack start point specified by the user to the attack end point is stored in the waveform memory as attack waveform data, and from the loop start point specified by the user to the loop end. Music waveform data up to the point is stored in the waveform memory as loop waveform data,
Calculating the pitch of the attack portion of the musical sound waveform signal obtained by reproducing the attack waveform data stored in the waveform memory by frequency analysis;
The pitch of the loop portion of the musical sound waveform signal reproduced by repeatedly reading out the loop waveform data stored in the waveform memory, the pitch of the loop portion obtained by frequency analysis, the number of samples from the loop start point to the loop end point Using the loop size representing and the predetermined sampling rate,
Wherein at least either one of the attack waveform data and the loop waveform data, the calculated resampled by changing the sampling rate according to the ratio of the pitch of the loop portion of the attack portion of the musical tone waveform signal A waveform data processing method characterized by the above.

Images