JP3454147B2 - Musical sound waveform analyzer, musical sound waveform analysis method, and computer-readable medium recording musical sound waveform analysis program - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a musical tone waveform analyzing apparatus and a musical tone waveform analyzing method for automatically extracting a harmonic component from a musical tone waveform, and a computer-readable medium recording a musical tone waveform analyzing program.

[0002]

2. Description of the Related Art Conventionally, the following processing has been performed in the technical field of tone analysis / synthesis for analyzing the properties of musical tones and synthesizing tone waveforms. First, a time-series sampling waveform is multiplied by a window function to cut out a frame, and fast Fourier transform (FFT) processing is performed to obtain a frequency spectrum component. This frequency spectrum component is data on the frequency axis, that is, frequency data, phase data, and amplitude data, and all frequency positions (frequency values) forming a peak are detected from the amplitude data. Then, the window of the window function is moved to update the frame, and the above processing is repeated to detect the peak in each frame.

Here, generally, a plurality of peaks in each frame are detected corresponding to a plurality of frequencies, and these peaks are a fundamental tone of the original sampling waveform, a harmonic component corresponding to its overtone, and a noise component. It is detected corresponding to the side lobe of the FFT window function. However, the perceptual pitch of a musical tone waveform is obtained mainly by the fundamental tone and the low-order harmonic components, and the timbre of the musical tone is analyzed so that the timbre can be obtained mainly by the higher-order harmonic components. In order to synthesize them, it is necessary to extract the harmonic component showing the characteristic of the musical sound from the peak of each frame. Therefore, conventionally, based on a known fundamental tone of the original sampling waveform, a peak is picked as a harmonic component in the vicinity of a frequency component having an overtone relation of an integral multiple thereof.

[0004]

However, in the conventional method of picking in the vicinity of the frequency component having an integral multiple harmonic relationship with the fundamental tone, the original harmonic component may be well extracted depending on the type of musical tone. It may not be possible and noise components may be extracted. As such a musical tone, a frequency spectrum of a musical tone of a piano is shown in FIG. FIG. 1 shows each line spectrum (only frequency in this figure) of a plurality of consecutive frames with the horizontal axis representing the time axis corresponding to the plurality of consecutive frames and the vertical axis representing the frequency axis. That is, one vertical line corresponds to one frame, and the points on that line are the line spectrum of that frame. In addition,
This example shows the middle C sound of a piano (fundamental frequency is approximately 26
4 Hertz).

As shown in the figure, the peaks are detected as being dot-like for each frame, having short continuity between frames, and having long continuity between frames. Of these, particularly long continuations are detected. Those having characteristics (those whose locus appears in frequency) correspond to the fundamental and harmonics that are characteristic of musical tones. However, particularly higher harmonics on the high frequency side tend to deviate from an overtone position (frequency value) that is an integral multiple of the fundamental tone, and the magnitude of this deviation becomes more pronounced at higher frequencies. For example, in FIG. 1, the harmonic component (peak locus) corresponding to the 14th overtone of the musical tone to be analyzed is located approximately in the middle between the 14th overtone and the 15th overtone calculated based on the fundamental frequency. A noise component appears at the position of the 14th harmonic. Therefore, in the conventional method, a noise component or the like between actual harmonic components may be picked as a harmonic component. An object of the present invention is to accurately extract a harmonic component of a musical tone waveform.

[0006]

The musical tone waveform analyzer according to claim 1 of the present invention is based on the frequency spectrum component of the musical tone waveform and is based on the low-order harmonics on the low frequency side to the high-order harmonics on the high frequency side. A repeating unit for performing a process of sequentially extracting the frequencies of harmonic components from low-order harmonics to high-order harmonics, which is a musical tone waveform analyzer for sequentially extracting the frequencies of harmonic components by applying waves. At the kth harmonic
When extracting the frequency, the frequency FP _{k-1 of the k−} 1th harmonic
Is multiplied by k / (k-1) and the multiplied value is multiplied by the frequency of the kth harmonic.
Difference between the estimation means for estimating a number, and the frequency F _i of the peak trace to be evaluated and the frequency estimated in said estimation means
Is evaluated by a function Func that takes as an argument the value obtained by dividing
The following expression (a) obtained by multiplying the amplitude A _i of the target peak locus
Calculating the evaluation function H _ki of the evaluation function H _ki is maximum
And a detection unit that detects the peak of the amplitude with the frequency F _i of the order i as the frequency of the k-th overtone . H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a)

In the musical tone waveform analyzing apparatus according to claim 1 configured as described above, the repeating means performs a process of sequentially extracting the frequencies of the harmonic components from the low order harmonics to the high order harmonics. The estimating means is the above repeating means.
When extracting the frequency of the kth harmonic, the frequency of the k-1th harmonic
The number FP _k-1 is multiplied by k / (k-1) and the multiplication value is multiplied by the k th
It is estimated as the frequency of the overtone, and the difference between the frequency estimated by the estimation means and the frequency F _{i of the} peak locus to be evaluated is calculated.
To the function Func that takes the value divided by the fundamental tone FB as an argument,
Of the following equation (a), which is obtained by multiplying the amplitude A _{i of the} peak locus as an elephant
The evaluation function H _ki is calculated, and the evaluation function H _ki is maximized next
The peak of the amplitude is detected with the frequency F _i of the number i as the frequency of the kth harmonic .

Since the frequency of the already extracted low-order harmonic component is affected by the deviation from the harmonic overtone position of an integer multiple of the fundamental tone, the high frequency estimated based on the frequency of the low-order harmonic component is detected. Estimated value of the frequency of the next harmonic component (FP _k-1
In addition, k / (k-1)) also takes into consideration the influence of the shift from the overtone position of an integer multiple of the fundamental tone. Therefore,
The frequency of the actual harmonic component deviated from the overtone position of an integer multiple of the fundamental tone ( the frequency of the order i at which the evaluation function H _ki is maximum).
F _i ) can be extracted. Also, the peak locus to be evaluated
High because of the use of the evaluation function H _ki in consideration of the amplitude A _i
The peak can be detected accurately.

According to a second aspect of the present invention, there is provided a method for analyzing a musical tone waveform according to the frequency spectrum component of the musical tone waveform, in which harmonic components are sequentially arranged from low-order harmonics on the low frequency side to high-order harmonics on the high frequency side. Is a musical tone waveform analyzing method for extracting the frequency of the k-th harmonic overtone .
The frequency FP _k−1 of the first harmonic is multiplied by k / (k−1)
The multiplication value is estimated as the frequency of the kth overtone, and the difference between the estimated frequency and the frequency F _{i of the} peak locus to be evaluated.
Is evaluated by a function Func that takes as an argument the value obtained by dividing
The following expression (a) obtained by multiplying the amplitude A _i of the target peak locus
Calculating the evaluation function H _ki of the evaluation function H _ki is maximum
It is characterized in that the peak of the amplitude is detected by using the frequency F _i of the order i as the frequency of the k-th overtone and the frequency of the harmonic component is extracted . H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a) According to the musical tone waveform analysis method of the second aspect configured as described above, the same operational effect as that of the first aspect can be obtained.

According to a third aspect of the present invention, the medium is composed of harmonic components sequentially from the low-order harmonics on the low frequency side to the high-order harmonics on the high frequency side based on the frequency spectrum component of the tone waveform. a recorded medium tone waveform analysis program for executing a process of extracting the frequency with computer, when extracting the frequency of the k-th harmonic, the k-1
The harmonic frequency FP _k-1 is multiplied by k / (k-1)
The calculated value is estimated as the frequency of the k-th overtone , and the estimated frequency and the frequency F _{i of the} peak locus to be evaluated are calculated.
Function Func whose argument is the value obtained by dividing the difference between
Is multiplied by the amplitude A _{i of the} peak locus to be evaluated.
The evaluation function H _ki of the equation (a) is calculated, and the evaluation function H _ki is the maximum.
Step, computer-readable recording the musical sound waveform analysis program for the execution of detecting the peak of <br/> amplitude frequency F _i as the frequency of the k harmonic of order i to be large
It is a medium. H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a) By executing the musical tone waveform analysis program recorded on the medium of the third aspect, the same effect as that of the first aspect can be obtained.

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below. FIG. 3 is a block diagram of a musical tone waveform analyzer according to an embodiment of the present invention, and shows a case where the musical tone waveform analyzer is configured by a personal computer including the CPU system 1. The program memory 2 is a hard disk device or a CD-ROM
It is a device or other external storage device, and stores a musical tone waveform analysis program described later. The data memory 3 is a RAM or the like, and stores musical tone waveform data to be analyzed and analysis results. Also, the input device 4
The performance operator 5 is a keyboard or the like, and the display 6 is a CRT or a liquid crystal display. Further, the network interface 7 is an interface connected to various networks 10 such as MIDI, LAN, and telephone lines.

The tone synthesizer 8 is a tone generator board or tone generator device having various LSI chips mounted thereon, and various registers (FR) which are interfaces to the CPU system 1.
CPU interface 81 with DATA register and RDATA register, multiple sine wave generation circuits (SWG1 to SWG1)
WGn) having a sinusoidal waveform calculation section 82, a residual waveform generation section 83 for generating a residual waveform described later, and an output of the sinusoidal waveform generation section 82 and an output of the residual waveform generation section 83 are combined. It is composed of a mixer and a synthesizing converter 84 including a DAC (digital-analog converter) that converts the mixer output into an analog signal.

With the above configuration, the CPU system 1 performs the waveform analysis processing by Fourier analysis on the designated musical tone waveform data in the data memory 3 based on the waveform analysis program stored in the program memory 2, and the frequency of each frame. Obtain basic analysis data ANLDATA such as data, amplitude data, phase data (frame data). In addition, the line spectrum component is Fourier-synthesized based on the frame data of the basic analysis data ANLDATA, and the difference between the original waveform data and the Fourier-synthesized waveform data is obtained as residual waveform data RDATA. And
By outputting the frame data of the basic analysis data ANLDATA and the residual waveform data RDATA stored in the data memory 3 to the musical tone synthesizer 8, the musical tone synthesizer 8 synthesizes musical tone waveforms.

That is, the musical tone synthesizing section 8 generates a Fourier synthetic wave by synthesizing the sine waves of the line spectrum in the sine wave waveform generating section 82 based on the frame data FRDATA set in the FRDATA register of the CPU interface 81. , The residual waveform generation unit 83 generates a residual waveform based on the residual waveform data RDATA set in the RDATA register of the CPU interface 81, and the synthesis conversion unit 84 synthesizes the Fourier synthetic wave and the residual waveform to generate a digital signal.・ Analog conversion is performed to generate a tone signal.

Next, the outline of the waveform analysis processing in the embodiment will be described. The original waveform data to be analyzed is time-series sampling data in which the amplitude of a musical tone waveform is quantized.For this original waveform data, a Fourier analysis process is performed on each frame cut out by moving the window function on the time axis. The frequency spectrum data is obtained for each time-series frame. The data of the frequency spectrum is data of frequency, phase and amplitude as data on the frequency axis. Then, the peak of the amplitude in this frequency spectrum is extracted to obtain, for example, the line spectrum component as shown in FIG. 1 (the line spectrum of one frame is composed of points scattered on the vertical axis). Although only the frequency data (vertical axis value) is shown in FIG. 1, the phase and amplitude data corresponding to each point are also stored at the same time.

Next, a peak locus in which peaks are continuous between frames is detected from the line spectrum, and a locus selection process selects a peak locus corresponding to a harmonic component from the peak locus. At this time, although it is possible for a person to determine the peak locus corresponding to the harmonic component by looking at the display image as shown in FIG. 1, the frame data associated with each frame (ie, corresponding to a plurality of peaks). In the case of automatic detection from frequency, phase, and amplitude data, the noise component or the like is detected at a particularly high order if the judgment is made only by the harmonic overtone position of an integer multiple of the fundamental tone as in the conventional case.

Therefore, by a "harmonic picking process" which will be described later, first, the peak locus of the frequency of the fundamental tone is obtained from the pitch from the frame data corresponding to that frame, and then the peak locus of the second overtone is obtained. Then, the position (frequency value) of the peak locus on the higher order side is sequentially estimated from the obtained peak locus, and the peak locus corresponding to the higher-order harmonic component is obtained.

FIG. 4 is a flowchart of the analysis process in the embodiment, and FIGS. 5 to 7 are flowcharts of the harmonic picking process in the embodiment. The process of the embodiment will be described based on each flowchart. In the following description and each flowchart, each register and the like are labeled with the following label, and the stored contents are labeled with the same label unless otherwise specified. In addition, in the flowchart, subscripts are written in half-width characters for easy viewing.

M: register of frame index i: register of peak order index N: register of maximum value of peak order i (number of peak loci-1) M: register of maximum value of frame index H _1i : peak locus of fundamental tone Register H _2i of the evaluation value for the peak order i at the time of detection: Register H _3i of the evaluation value for the peak order i when the peak locus of the second overtone is detected H _3i : For the peak order i at the peak locus of the third overtone Register of evaluation value H _ki : Register of evaluation value for peak order i at peak locus detection of k-th harmonic FB: Register of pitch of fundamental tone FP ₁ : Register of frequency of fundamental tone detected FP ₂ : Detected second overtone register frequency FP _3: register FP _k of the frequency of the third harmonic detected: the frequency of the k-th harmonic of the detected register a _i: Register of the amplitude of the peak order i in frame

In the analysis process of FIG. 4, in step S1, the process of designating the original waveform data to be analyzed, the process of setting the pitch when the pitch of the original waveform data is known in advance, the harmonics up to the harmonic overtone. Various preparatory processes are performed, such as the process of determining analysis specifications such as whether to analyze. If the pitch of the original waveform data is not known, step S
At 1 ', pitch detection processing is performed. Next, in step S2, the window function is multiplied to the original waveform data to cut out a frame, the SFFT process (short-time fast Fourier analysis) is performed in step S3, and the entire section of the original waveform data is analyzed in step S4. Is completed. If not completed, window function time shift (frame time shift) is performed in step S5, and the process returns to step S2.

When the analysis of all sections is completed, step S6
In step S7, spectrum peak picking processing for detecting the peak of the amplitude of the frequency spectrum in each frame data is performed. In step S7, peak locus detection processing for detecting a continuous portion of the peaks between frames is performed.
The trajectory selection processing is performed with. In this trajectory selection process, the harmonic picking process of FIGS. 5 to 7 described later is performed.

Upon completion of the trajectory selection process in step S8, basic analysis data AN such as frame data (frequency data, amplitude data, phase data) of the fundamental to higher harmonic components
Since LDATA is obtained, basic analysis data is obtained in step S9.
ANLDATA is stored in the data memory 3, and the process of re-synthesizing the musical sound is performed in step S10. In this recombining process, the component waveform of the line spectrum is generated for each frame time based on the frame data (frequency, phase, and amplitude data) of the basic analysis data, and Fourier combining is performed. Next, in step S11, the difference between the original waveform data and the re-synthesized waveform data is calculated to obtain residual waveform data RDATA. In step S12, this residual waveform data RDATA is stored in the data memory 3, and the analysis process ends. .

Through the above processing, basic analysis data ANLDATA and residual waveform data RDATA, which are the analysis results of the original waveform data, are obtained. And this basic analysis data ANLD
The ATA frame data FRDATA is used as the FRDATA register and RDATA register (CPU interface 8
By setting each of 1), a musical sound is reproduced by the musical sound synthesizer 8 as described above.

Next, the harmonic picking process of FIGS. 5 to 7 will be described. First, in step S20, "1" is set to the frame index m, and then step S2
At 1, the peak order index i is reset to “0”,
By the determination in step S23 and the increment of i in step S24, the processing in step S22 is performed for all orders of i = 0, 1, ..., N. In step S22, FIG.
Function Func [], fundamental pitch FB, peak order i
The evaluation function H _1i of the following equation (1) is calculated from the frequency F _i and the amplitude A _i of the peak locus of H _1i = Func [| FB−F _i | / FB] · A _i (1) Note that the variable F of the function Func [| F−x | / FB] is the pitch FB of the fundamental tone or the frequency of the estimated peak locus. FP ₁ ~
FP _k−1 is substituted, and the frequency F _{i of the} peak locus to be evaluated is substituted for the variable x.

The above calculation of H _1i is performed for all orders, and when i = N in step S23, step S25
And determine the frequency Fi of order i that maximizes H _{1i at
It} is stored as ₁ , and the process proceeds to step S26. That is, this FP ₁ becomes a fundamental frequency determined based on the line spectrum data from the pitch FB of the fundamental tone.

Steps S26 and thereafter are processings for determining the frequencies of higher harmonics of the second and subsequent harmonics.
Similar to 21 to S25, the peak order index i is reset (steps S26, S31, S36), i is incremented (steps S29, S34, S39), and N is determined (steps S28, S33, S38) to evaluate. Function calculation (steps S27, S32, S3
7) is calculated for all orders.

That is, in step S27, using the fundamental frequency FP ₁ determined in step S25, the following equation (2)
The evaluation function H _2i of H _2i = Func [| 2 · FP ₁ −F _i | / FB] · A _i (2) Then, H _2i is calculated for all orders, and when i = N in step S 28, in step S 30, The frequency F _{i of the} order i that maximizes H _2i is determined and stored as FP ₂ , and the process proceeds to step S31 in FIG. That is, this F
P ₂ is a harmonic frequency corresponding to the second overtone determined from the fundamental frequency FP ₁ based on the data of the line spectrum.

In step S32 of FIG. 6, the evaluation function H _3i of the following equation (3) is calculated using the frequency FP ₂ of the harmonic corresponding to the second overtone. H _3i = Func [| 3 · FP ₂ / 2-F _i | / FB] · A _i (3) Then, H _3i is calculated for all orders, and when i = N in step S 33, step S 35 Then, the frequency F _{i of the} order i that maximizes H _3i is determined and stored as FP ₃ , and the process proceeds to step S36. That is, this FP ₃ becomes the frequency of the harmonic corresponding to the third overtone determined from the frequency FP _{2 of the} harmonic corresponding to the second overtone based on the data of the line spectrum.

Thereafter, the same processing is sequentially performed for the higher harmonics, and the processing for the kth harmonic is generally performed in step S37.
Done in. That is, in step S37, using the frequency FP _k−1 of the harmonic corresponding to the k−1 th harmonic, the following equation (4)
Of the evaluation function H _ki of H _ki = Func [｜ k ・ FP
_k-1 / (k-1) -F _i | / FB] · A _i (4) Then, the calculation of H _ki is calculated for all orders, and step S3
When i = N in 8, the frequency F _{i of the} order i that maximizes H _ki is determined and stored as FP _k in step S40, and the process proceeds to step S41 in FIG. That is, this FP
_k is the frequency of the harmonic corresponding to the k-th overtone determined from the frequency FP _k-1 of the harmonic corresponding to the k−1-th overtone based on the data of the line spectrum.

Here, the estimated value of the higher harmonic frequency estimated from the lower harmonic is, for example, in equation (4), "k · FP _k-1 /" in the first term in the absolute value symbol. Corresponding to (k-1) "and detecting the peak of the amplitude in the vicinity of the estimated frequency is" processing to determine the frequency F _{i of the} order i that maximizes H _ki and set it as FP _k ". It corresponds to.
The flowcharts of FIGS. 5 and 6 show the process for the kth overtone as a generalized process that can be applied after the second overtone after the process of the fundamental frequency, the second overtone, and the third overtone. In the program, the process for the kth overtone may be applied to the second and subsequent overtones by a loop process using an appropriate variable.

The above processing is for one frame. In step S41 of FIG. 7, it is determined whether "m = M", that is, whether the frame index m reaches the maximum value M, If not "m = M", step S
At 42, the frame index m is incremented, the process returns to step S21 (FIG. 5), and the process for the next frame is performed. If "m = M", in step S43, a component of each order k whose frequency is [FP _k ] _m (m = 1 to M) is selected for each frame as the peak of the kth overtone. Then, in step S44, the FRDATA editing process is performed, such as deleting the line spectrum data deviating from the peak of the overtone in the frame data FRDATA of each frame, and the process ends.

FIG. 2 shows the frequency data of the frame data FRDATA when the harmonic picking processing is applied to the line spectrum components of FIG. 1 and the frame data FR.
It is a figure which shows the waveform which carried out the Fourier synthesis based on DATA.
Note that, in FIG. 2, the magnification on the time axis (horizontal axis) is enlarged to twice that of FIG. As the figure shows, the higher the frequency,
That is, it is possible to detect a peak locus of a harmonic component of a musical tone having a characteristic that higher harmonic overtones deviate from a frequency position that is a simple integer multiple of the fundamental frequency, and unnecessary noise components can be removed. .

In the above embodiment, the process of re-synthesizing the musical tone based on the frame data in step S10 is performed by software. However, as shown by the broken line in the musical tone synthesizing section 8 in FIG. FRDATA is a musical sound synthesizer 8
Alternatively, the sine wave waveform calculation unit 82 may output a waveform that has been Fourier-combined, and the output may be fetched to obtain the residual waveform data RDATA by calculating the difference from the original waveform data in step S11.

Further, in the above embodiment, the synthetic sound is generated by setting the frame data FRDATA and the residual waveform data RDATA residual in the musical sound synthesizing section 8, but a flow chart is shown in FIG. Thus, it may be processed by software. In the synthetic sound generation process of FIG. 9, first, in step S50, it is monitored whether or not there is a sounding instruction by the performance operator 5, etc., and if there is a sounding instruction, the processing of step S51 and thereafter is performed. In step S51, basic analysis data ANLDATA
A line spectrum component waveform is generated and synthesized for each frame time based on the frame data FRDATA of
P.wav, and in step S52, residual waveform data RD
Based on ATA, the residual waveform corresponding to the current frame or the time from the start of sound generation is reproduced and stored in the register RS.wav. Then, in step S53, RS.wav and RS.wav are added and output as a musical tone waveform, and in step S54, it is determined whether or not the sound generation ends. If the pronunciation has not ended, it is determined in step S55 whether it is the frame update timing.
Returning to step 1, if it is the frame update timing, the frame is updated in step S56 and the process returns to step S51. If the sound generation ends, the mute processing is performed in step S57 and the processing ends. Note that the frequency of each line spectrum component of the frame data FRDATA may be controlled in order to change the pitch of the generated musical sound.

In the above embodiment, the evaluation function H _1i is calculated for all peak orders when the fundamental frequency FP ₁ is calculated. However, in order to speed up the calculation, the peak clearly showing the fundamental frequency is calculated. It is not necessary to calculate the peaks that are determined not to be. For example, only the peaks within the range of FB / 2 <F _i <FB · 2 may be calculated. Similarly, when calculating the evaluation function H _ki , generally, only the peaks falling within the range of FP _k−1 / <F _i <(k + 1) · FP _k−1 / (k−1) need be calculated.

Further, since there may be waveform data in which halftones in the middle are missing, such as a waveform having only odd-numbered harmonics, the minimum value of the evaluation function H _ki is determined,
If the value of the maximum evaluation function H _ki is less than or equal to the minimum value, the kth harmonic should not be present. Then, as the estimated value of the frequency of the (k + 1) th overtone when there is no kth overtone,
(K + 1) · F using the frequency FP _k−1 of the k−1 th harmonic
Let P _k-1 / (k-1) be the estimated value. In the case of a waveform with no fundamental tone, the peak locus may be evaluated from the order above that, for example, the second overtone. In this case the second
The estimated value of the overtone frequency is 2 · FB.

There is also a waveform that includes subharmonics whose frequency is lower than the pitch. For example, even in the peak data of the musical tone of the piano shown in FIG. 1, a peak locus appears near 0.5 · BF (BF is the pitch). In such a case, if picking including subharmonic is desired, FB of the above processing
Should be replaced with 0.5 · FB. That is, it may be evaluated that the fundamental tone is 0.5 FB. In this case, the 3rd, 5th, and 7th overtones are not found, but as described above, the estimated value of the frequency of the k + 1th overtone in the absence of the kth overtone is (k + 1) · FP _k−1 / (k− This can be dealt with by the process 1).

As described above, in order to estimate the frequency of the higher-order harmonic, it is not limited to the estimation based on the order below the first order, but any order lower than the estimated order may be used. Note that it is preferable to make an estimation based on the order that is one order lower.

In the embodiment, the waveforms are synthesized and the musical sounds are generated after the waveforms are analyzed. However, the present invention relates to the waveform analysis. It is not limited. Further, the residual waveform data may not be generated.

In the above embodiment, the case where the musical tone waveform analysis program is stored in the program memory 2 in advance has been described, but the present invention is not limited to this, and may be performed as follows. For example, the musical tone waveform analysis program is recorded in a CD-ROM, and the musical tone waveform analysis program is loaded from the CD-ROM device to the hard disk. And CPU
The system 1 develops the musical tone waveform analysis program of the hard disk in the RAM or the like, and controls the operation of the musical tone waveform analysis as in the above-described embodiment based on the program of the RAM. As a result, the CPU can be caused to perform the same operation as when the musical tone waveform analysis program is stored in the program memory 2. In this way, new installation, addition, or version upgrade of the musical tone waveform analysis program can be easily performed. Also, a floppy disk,
The musical tone waveform analysis program may be recorded on a magnetic disk (MO) or the like and supplied to the RAM or the hard disk.

The network interface 7 may be used to download the musical tone waveform analysis program. In this case, for example, by connecting to a network 10 such as a LAN (local area network), the Internet, or a telephone line, and receiving the distribution of the musical tone waveform analysis program from the server computer via the network 10, it is recorded on the hard disk. Then the download is completed.

The present invention is not limited to the personal computer as in the above embodiment, but may be a keyboard type electronic musical instrument, sound source device,
A sequencer, an effector, and the like may be separate devices, and each device may be connected using communication means such as MIDI or various networks.

The medium in which the musical tone waveform analysis program as described in the above embodiment is recorded, that is, RO
The storage device of the delivery destination such as M, RAM, hard disk, CD-ROM, magneto-optical disk, DVD (digital multipurpose disk) or network server computer corresponds to the medium recording the musical tone waveform analysis program of claim 3 of the present invention. To do.

[0044]

As described above, according to the first aspect of the present invention.
According to the musical tone waveform analyzing apparatus described above, the musical tone waveform analyzing method according to claim 2, or the musical tone waveform analyzing program recorded on the medium according to claim 3, the frequency of the k-th harmonic is
When extracting the number , k is added to the frequency FP _k-1 of the k−1 th harmonic.
/ (K-1) and multiply the product by the frequency of the kth harmonic.
Peak was estimated, it is evaluated and frequencies the estimated
Subtract the value obtained by dividing the difference between the locus frequency F _i and the fundamental frequency FB.
The function Func that is a number, and the amplitude of the peak locus to be evaluated
The evaluation function H _ki multiplied by A _i is calculated, and the evaluation function H _ki is calculated.
The frequency of the k-th harmonic frequency F _i of order i which maximizes but
Since the peak of the amplitude is detected as the frequency of the harmonic component, the frequency of the actual harmonic component that is deviated from the harmonic overtone position of an integer multiple of the fundamental tone can be extracted, and the harmonic component of the musical tone waveform can be accurately extracted. can do.

[Brief description of drawings]

FIG. 1 is a diagram showing a frequency spectrum of a musical sound of a piano for explaining an embodiment of the present invention and a conventional problem.

FIG. 2 is a diagram showing a result of performing the harmonic picking process of the embodiment on the spectral components of FIG. 1 and a synthesized waveform.

FIG. 3 is a block diagram of the musical tone waveform analyzer of the embodiment.

FIG. 4 is a flowchart of analysis processing according to the embodiment.

FIG. 5 is a part of a flowchart of harmonic picking processing in the embodiment.

FIG. 6 is another part of the flowchart of the harmonic picking process in the embodiment.

FIG. 7 is still another part of the flowchart of the harmonic picking process in the embodiment.

FIG. 8 is a diagram showing a function Func [] for calculating an evaluation function in the embodiment.

FIG. 9 is a flowchart of a synthetic tone generation process in the embodiment.

[Explanation of symbols]

1 ... CPU system, 2 ... program memory, 3 ... data memory, 8 ... tone synthesis section.

Claims (3)

(57) [Claims] 1. A musical tone waveform analyzer for sequentially extracting frequencies of harmonic components from low-order harmonics on the low frequency side to high-order harmonics on the high frequency side based on the frequency spectrum components of the musical tone waveform. A repeating means for sequentially extracting frequencies of harmonic components from low-order harmonics to high-order harmonics; and when extracting the frequency of the kth harmonic by the repeating means,
Multiply the frequency FP _k-1 of the k-1 th harmonic by k / (k-1)
To an estimation means for estimating a multiplication value as the frequency of the k-th harmonic, peak trajectories to be evaluated and the frequency estimated in said estimation means
The argument is the value obtained by dividing the difference from the trace frequency F _i by the fundamental tone FB.
To the function Func that is the amplitude A of the peak locus to be evaluated
The evaluation function H _ki of the following expression (a) multiplied by _i is calculated, and the evaluation
The frequency F _{i of} order i that maximizes the valence function H _ki is set to the kth harmonic
And a detecting means for detecting the peak of the amplitude as the frequency of the musical tone waveform analyzer. H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a) 2. A musical tone waveform analysis method for sequentially extracting frequencies of harmonic components from low-order harmonics on the low frequency side to high-order harmonics on the high frequency side based on frequency spectrum components of the musical tone waveform. , when extracting the frequency of the k-th harmonic, frequency of the k-1 overtone
The number FP _k-1 is multiplied by k / (k-1) and the multiplication value is multiplied by the k th
Estimated as the overtone frequency, the estimated frequency and the frequency of the peak locus to be evaluated
A function that takes as an argument the value obtained by dividing the difference from F _i by the fundamental tone FB
Multiply Func by the amplitude A _{i of the} peak locus to be evaluated.
And it calculates the evaluation function H _ki of the formula (a), the evaluation function H _ki
The frequency of the k-th harmonic frequency F _i of order i which maximizes but
Then, the peak of the amplitude is detected, and the frequency of the harmonic component is extracted. H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a) 3. A computer executes a process of sequentially extracting the frequencies of the harmonic components from the low-order harmonics on the low frequency side to the high-order harmonics on the high frequency side based on the frequency spectrum components of the musical tone waveform. a recorded medium tone waveform analysis program, when extracting the frequency of the k-th harmonic, frequency of the k-1 overtone
The number FP _k-1 is multiplied by k / (k-1) and the multiplication value is multiplied by the k th
The frequency of the overtone is estimated , and the difference between the estimated frequency and the frequency F _{i of the} peak locus to be evaluated is calculated.
To the function Func that takes the value divided by the fundamental tone FB as an argument,
Of the following equation (a), which is obtained by multiplying the amplitude A _{i of the} peak locus as an elephant
The evaluation function H _ki is calculated, and the evaluation function H _ki is maximized next
The number i of the frequency F _i of steps, a computer readable medium storing tone waveform analysis program that performs for detecting a peak amplitude as the frequency of the k-th harmonic. H _ki = Func [│k · FP _k-1 / (k-1) -F _i ｜ / F
B] · A _i (a)