JP3767236B2 - Musical sound waveform analyzer, musical sound waveform analysis method, and computer-readable recording medium recording a musical sound waveform analysis program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention performs a Fourier transform process on a musical sound waveform to be analyzed to obtain a frequency spectrum of time-series frames, detects peak points of amplitude data of analysis data in each frame, and continues in the time axis direction of the frame. A musical tone waveform analyzer, a musical tone waveform analysis method, and a musical tone waveform analysis program for tracking the locus of the peak point are recorded. Computer readable The present invention relates to a recording medium.
[0002]
[Prior art]
Conventionally, the following processing is performed in the technical field of musical sound analysis and (re) synthesis (Analysis & (Re) Synthesis) for analyzing the characteristics of musical sounds and synthesizing musical sound 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 frequency spectrum analysis data. The analysis data of the frequency spectrum includes data on the frequency axis, that is, frequency data, amplitude data, and phase data, and all peak points (frequency values) that can be regarded as peaks of the amplitude data are detected. Then, the window of the window function is moved to update the frame, the above processing is repeated, and the peak point of the amplitude data of the frequency spectrum in each frame is detected.
[0003]
Here, a plurality of peak points in each frame are generally detected corresponding to a plurality of frequencies, and these peak points are a fundamental tone of the original sampling waveform, a harmonic component corresponding to its harmonic, a noise component, an FFT, and the like. Are detected corresponding to the side lobes of the window function. However, the audible pitch of the musical tone is obtained mainly from the fundamental tone and the lower harmonic components, and the characteristics of the musical tone are analyzed and the musical sound waveform is synthesized so that the timbre is obtained mainly from the higher harmonic components. In order to achieve this, it is important to extract peak points corresponding to fundamental and overtones. Since the peak corresponding to the fundamental tone or the harmonic overtone forms a long trajectory continuous in the time axis direction of the frame in the FFT, it is required to track the trajectory of the peak point in these musical sound waveform analysis processes.
[0004]
FIG. 7 is a diagram illustrating an example of the peak of the amplitude data of the frequency spectrum obtained as a result of the analysis by FFT. The example shown in this figure is an analysis of a piano C4 key (center c) sound (fundamental frequency is approximately 261.63 Hz), the horizontal axis is time (unit is ms), and the vertical axis is frequency (unit is Hz), and the peak points detected for each frame corresponding to the time window are represented by dots. Moreover, the horizontal thin line represents the fundamental frequency and its theoretical overtone frequency. In this figure, all the points exist independently, and you can see the line drawing the trajectory in the figure, but the information that this trajectory is connected is not yet obtained, each frame Only the information on which frequency is the peak at each time is obtained. In addition, it is considered that the part other than the part that appears as a long trajectory is noise or that the side lobe of the FFT window function is displayed.
[0005]
In order to extract a peak point having a locus from such peak point data, generally, a component having a harmonic relationship is picked in each frame. That is, it is determined that the peak point closest to the known fundamental frequency FB and its integral multiples 2 × FB, 3 × FB,... Is the fundamental tone or harmonics, and harmonic data is picked up for each frame. The method is taken. Then, it is determined that the peak points detected with respect to the fundamental tone or harmonics between frames form one locus.
[0006]
[Problems to be solved by the invention]
FIG. 8 is an enlarged view of the fundamental tone in FIG. 7 around 15 msec. In this figure, when the tracking direction of the locus of the peak point is from right to left, it is natural that the peak point is connected in the direction of the arrow at the frame time A. However, in the above-described conventional method, for example, when looking at the fundamental frequency (261.63 Hz in this case), the following is performed under the condition that the peak point closest to the fundamental frequency is selected in each frame. The following problems arise. For example, in each frame from the right side of FIG. 8 to frame time A, there is one peak in the vicinity of the fundamental frequency, but at frame time A-1 next to frame time A, peak point b is higher than peak point a. It occurs close to the fundamental frequency. For this reason, when the frame time A changes from A to A-1, the locus is connected to the peak point b, and the locus of the peak point may not be tracked well.
[0007]
It is an object of the present invention to make it possible to trace the locus of peak points continuous in the time axis direction of a frame with high accuracy with respect to the peak points of each frequency spectrum in a time-series frame of an analysis target musical sound waveform.
[0008]
[Means for Solving the Problems]
The musical sound waveform analyzer according to claim 1 of the present invention performs a Fourier transform process on the musical sound waveform to be analyzed to obtain a frequency spectrum of a time-series frame, and detects a peak point of amplitude data of analysis data in each frame. In the musical tone waveform analyzer for tracking the locus of the peak point that is continuous in the time axis direction of the frame, the cell space is composed of cells defined by frequency coordinates and time coordinates, and at least the time coordinates are the frame. The unit is the time interval that divides Storage means in which a cell space defined as coordinates is set; An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is forward or backward in time with respect to the center cell and the center cell. A calculation unit that uses a template in which a coefficient is assigned to each cell, superimposes the center of the template on each of the peak points, and multiplies the amplitude coefficient of each cell of the template at that time by the amplitude of the peak point; Accumulating means for accumulating each multiplication value obtained by the computing means for each cell in the cell space; Peak tracking means for tracking the locus of the peak point by selecting the cell following the extreme value of the amplitude accumulated in each cell in the time axis direction of the frame. And
[0009]
According to the musical tone waveform analyzer of claim 1 configured as described above, the cell space set in the storage means is composed of cells defined by frequency coordinates and time coordinates, and at least the time coordinates are the frames. The unit is the time interval that divides It is defined as coordinates. The calculation means is An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is forward or backward in time with respect to the center cell and the center cell. A template in which a coefficient is assigned to each cell is used, the center of the template is superimposed on each of the peak points, and the amplitude coefficient of each cell of the template at that time is multiplied by the amplitude of the peak point. The accumulating means accumulates each multiplication value obtained by the calculating means for each cell in the cell space. This accumulated amplitude tends to have an extreme value (for example, a local maximum value) in a cell on the path having the shortest distance between adjacent peak points. Therefore, if the peak tracking means tracks the peak point trajectory by selecting the cell following the extreme value of the weight accumulated in each cell in the time axis direction of the frame, the peak point trajectory has high accuracy. Can be tracked at.
[0010]
The musical sound waveform analysis method according to claim 2 of the present invention obtains a frequency spectrum of a time-series frame by performing a Fourier transform process on the musical sound waveform to be analyzed, and detects a peak point of the amplitude data of the analysis data in each frame. In the musical sound waveform analysis method for tracking the trajectory of the peak point continuous in the time axis direction of the frame, a cell space consisting of cells defined by frequency coordinates and time coordinates, and at least the time coordinates of the frame The unit is the time interval that divides Using cell space defined as coordinates, An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is forward or backward in time with respect to the center cell and the center cell. Using a template in which a coefficient is assigned to each cell, the center of the template is superimposed on each of the peak points, the amplitude coefficient of each cell of the template at that time is multiplied by the amplitude of the peak point, and the multiplication is performed. Each obtained multiplication value is accumulated for each cell on the cell space, The locus of the peak point is traced by selecting the cell following the extreme value of the amplitude accumulated in each cell in the time axis direction of the frame.
[0011]
According to the musical sound waveform analysis method of claim 2 configured as described above, as in the case of claim 1, in the cell space, the amplitude pole accumulated in each cell in the time axis direction of the frame. Since the cell is selected according to the value and the locus of the peak point is traced, the locus of the peak point can be traced with high accuracy.
[0012]
Claim 3 of the present invention Computer readable The recording medium performs a Fourier transform process on the musical tone waveform to be analyzed to obtain a frequency spectrum of a time-series frame, detects a peak point of the amplitude data of the analysis data in each frame, and continuously in the time axis direction of the frame Recorded a musical sound waveform analysis program for executing a process of tracking the locus of the peak point on a computer Computer readable A recording medium, which is a cell space composed of cells defined by frequency coordinates and time coordinates, and at least the time coordinates of the frame The unit is the time interval that divides Using the cell space defined as coordinates, associating the detected peak point with a cell in the cell space according to the frequency and frame of the peak point, and from the peak point to a cell in the vicinity of the peak point of Defined in the cell space Setting the amplitude of the peak point with a weight according to the distance and accumulating the amplitude set for each cell; An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is forward or backward in time with respect to the center cell and the center cell. Using a template in which a coefficient is assigned to each cell, superimposing the center of the template on each of the peak points, and multiplying the amplitude coefficient of each cell of the template at that time by the amplitude of the peak point; Accumulating each multiplication value obtained by multiplication for each cell in the cell space; In the time axis direction of the frame, following the extreme value of the amplitude accumulated in each cell, the step of tracking the peak point by selecting the cell and tracking the peak point trajectory is executed by a computer According to the execution of the musical sound waveform analysis program recorded on the recording medium according to the third aspect, the same effects as those of the first and second aspects can be obtained.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a block diagram of a musical sound waveform analyzer according to an embodiment of the present invention, and shows a case where the musical sound wave analyzer is constituted by a personal computer equipped with a CPU 1. The program memory 2 is a hard disk device, a CD-ROM device, or other external storage device, and stores a musical sound waveform analysis program described later. The data memory 3 is a RAM or the like, and stores musical sound waveform data to be analyzed, analysis results, and the like. The input device 4 and the performance operator 5 are a keyboard or the like, and the display 6 is a CRT or a liquid crystal display. The tone synthesis unit 7 is a tone generator board or tone generator equipped with various LSI chips for synthesizing a tone based on the analysis result. Further, the network interface 8 is an interface for connecting to various networks 9 such as MIDI, LAN, telephone line and the like.
[0014]
Based on the waveform analysis program stored in, for example, a hard disk device of the program memory 2, the CPU 1 performs waveform analysis processing, which will be described later, on the musical tone waveform data designated in the data memory 3, and the frequency spectrum in each frame of the Fourier analysis processing The peak point of is detected. And the peak point which makes a locus | trajectory between frames is extracted, and the analysis data of the peak point in each frame which makes this locus | trajectory are obtained as an analysis result.
[0015]
Note that the CPU 1 uses, for each frame, frequency data, amplitude data, and phase data, which are analysis data of the peak points forming the trajectory, based on the trajectory of the peak points obtained as a result of the waveform analysis described later. The musical sound synthesizing unit 7 performs Fourier synthesis by outputting them together to the musical sound synthesizing unit 7. Further, the difference between the original waveform data and the Fourier synthesized waveform data is obtained as residual waveform data. The residual waveform data and the Fourier-synthesized waveform data are re-synthesized by the tone synthesizing unit 7 and digital-analog converted to be generated as a tone signal. Also, desired waveform data can be obtained by modifying the analysis data and residual waveform data output to the musical tone synthesis unit 7.
[0016]
FIG. 2 is a flowchart of the waveform analysis process executed by the CPU 1 in the embodiment, and the operation of the waveform analysis process in the embodiment will be described with reference to FIG. FIG. 6 is a diagram showing an example of a time waveform of the sound of the C4 key of the same piano as in FIG. 7, and a case where peak tracking is performed using this waveform as an analysis target will be described. The waveform data of the time waveform is time-series sampling data obtained by sampling and quantizing the amplitude of the musical sound waveform at a sampling frequency of 44.1 kHz, for example, and is stored in the data memory 3.
[0017]
When the waveform analysis process is started, first, frame analysis of the analysis target sound waveform is performed in step S1. In this frame analysis processing, first, in order to perform Fourier analysis processing, initial settings such as setting of analysis range, window size, FFT size, and HOP size for moving the window function along the time axis are performed. In general, it is appropriate that the FFT window size is an integral multiple of the fundamental period of the waveform, for example, 8 times. The sound of the C4 key shown in FIG. 6 has a fundamental frequency of 261.63 Hz, and the number of sampling points included in one period is 168.6 because the sampling frequency is 44.1 kHz. Therefore, the window size is 8 times 1348 points (rounded off after the decimal point).
[0018]
Furthermore, the FFT size must be a power of 2, and the larger the FFT size, the better the frequency resolution. However, even if it is too large, it takes a long time to calculate. Therefore, the minimum number of powers of 2 exceeding the window size is selected, and 2048 points are set. Furthermore, the HOP size is set to 1/8 of the basic period and 21 points. Under these conditions, Fourier analysis processing is performed on each of the frames extracted by moving the window function on the time axis with respect to the waveform data shown in FIG. 6 to obtain frequency spectrum analysis data for each time-series frame. . The analysis data of the frequency spectrum is frequency data, amplitude data, and phase data as data on the frequency axis.
[0019]
Next, in step S2, the peak of the amplitude data of the analysis data of each frame in the frequency spectrum is extracted, for example, the line spectrum component as shown in FIG. 7 (the line spectrum of one frame is discrete on the vertical axis). Configured.) Although only frequency data (value on the vertical axis) corresponding to each peak point is shown in FIG. 7, amplitude data and phase data corresponding to each peak point are also stored in the data memory 3 at the same time.
[0020]
Next, in step S3, the frame-freq plane obtained by subdividing the frame time and frequency is set as a cell space in a predetermined area of the data memory 3, and the peak point detected on this frame-freq plane is set. Deploy. FIG. 3 is a conceptual diagram showing a state in which peak points are arranged on the frame-freq plane. The frame-freq plane is composed of cells arranged in a matrix defined by time coordinates and frequency coordinates. The time coordinate is set to four times the resolution of the frame time (the resolution at which the HOP size is 1), and the frequency coordinate is set to 900 Hz / 120 = 7.5 Hz as one unit. The numerical value of one unit of the frequency coordinate is set to about 1/32 with respect to about 261.63 Hz which is the frequency of the fundamental sound of the analysis target sound waveform. (And frequency interval between overtones). The same thing is almost possible with 1/16. In the figure, for convenience, frequency coordinates are represented by alphabets “A, B, C,...”, Time coordinates are represented by numbers “1, 2, 3,...”, And a cell is designated in the following description. To do this, a coordinate display such as [A, 1] ([frequency coordinate, time coordinate]) is used. Further, since there is no possibility of confusion with the frame of the Fourier analysis process, the time coordinate is also referred to as a “frame” as appropriate. In FIG. 3, peak points are arranged in cells in which numerical values other than “0” are entered (for example, “34” [N, 5] cell, “35” [M, 9] cell, etc.). Each numerical value indicates the amplitude of the peak point in dB (decibel).
[0021]
Next, in step S4, the amplitude coefficient template shown in FIG. 4 is applied to each peak point, and the amplitude coefficient is converted into the amplitude of the peak point with respect to the cell on the frame-freq plane corresponding to each cell of the amplitude coefficient template. A value obtained by multiplying the amplitude coefficient of the template cell is set, and the set multiplication value is accumulated in each cell on the frame-freq plane. As a result, an accumulation result on the frame-freq plane as shown in FIG. 5 is obtained.
[0022]
Hereinafter, a specific example of the process of step 4 will be described based on FIGS. 3, 4, and 5. The amplitude coefficient template of FIG. 4 includes five frequency coordinates indicated by alphabets “a, b, c, d, e” and numbers “1, 2, 3, 4, 5, 6, 7, 8, 9”. Is a template in which amplitude coefficients are arranged in a 5 × 9 matrix cell defined by the nine time coordinates. The amplitude coefficient is set to “1” at the center cell, and for example, “1/2 → 1/4 → 1/8 → 1/16” or “1/8 → 1/16 → 1” in the surrounding cells. It is set to a value that decreases with increasing distance from the center, such as / 32 → 1/64 ″. As in the case of the frame-freq plane, a coordinate display such as <a, 1>(<frequency coordinate, time coordinate>) is used when a cell of an amplitude coefficient template or its amplitude coefficient is designated.
[0023]
For example, the <c, 5> cell of the amplitude coefficient template in FIG. 4 is matched with the [N, 5] cell at the peak point of the frame-freq plane in FIG. e ”corresponds to the frequency coordinates“ L,..., P ”of the frame-freq plane, and the time coordinates“ 1,..., 9 ”of the amplitude template are changed to the time coordinates“ 1,. Correspondingly, the cell of the frame-freq plane is associated with the cell of the amplitude coefficient template. Next, on the frame-freq plane, the peak point cell is set with the amplitude of the peak point (in this example, “34”), and the other cells are peaked with the amplitude coefficient of the corresponding cell of the amplitude coefficient template. Multiply the amplitude of the point and set the multiplication value. For example, the multiplication value “17” obtained by multiplying “34” by the amplitude coefficient “1/2” of the cell of <c, 4> is set to the cell [N, 4], and [N, 3]. For the cell corresponding to the amplitude coefficient template, for example, a multiplication value “18” (rounded down to the decimal point) obtained by multiplying “34” by the amplitude coefficient “1/4” of the cell <c, 3> is set in Then, a process of setting a multiplication value of the amplitude value and the amplitude coefficient is performed.
[0024]
Further, the cell of the amplitude coefficient template is matched with the cell of the frame-freq plane and the cell of the amplitude coefficient template in the same manner as described above by matching the cell of <c, 5> of the amplitude coefficient template with the cell of [M, 9] at the peak point of the frame-freq plane. In addition, similarly to the above, processing for setting a multiplication value of the amplitude value and the amplitude coefficient is performed for the cells around [M, 9] at the peak point. Then, the process of associating the cells on the frame-freq plane with the cells of the amplitude coefficient template and setting the multiplication value of the amplitude value and the amplitude coefficient is executed for all peak points. As a result, a plurality of multiplication values corresponding to the amplitude value of the adjacent peak point and its amplitude coefficient are set in the cell on the frame-freq plane, and the multiplication value is accumulated in each cell. With the above processing, for example, an accumulation result other than 0 is obtained in a cell connected in the vicinity of the peak point on the frame-freq plane as shown in FIG. For example, the cell of [N, 7] in FIG. 5 is “12”, which corresponds to “1/4” of <c, 7> in “34” of the cell of [N, 5] in FIG. Is multiplied by “8” (rounded down) and “35” in the cell [M, 9] in FIG. 3 is multiplied by “1/8” in <d, 3> (4) ( It is an added value (accumulated value) with a decimal point.
[0025]
When the accumulated value as described above is obtained, the trajectory tracking process is performed in step 5 as follows. In this trace tracking process, an appropriate peak point is set as a base point, a cell in the frame time direction (time coordinate axis direction) is searched from the peak point on the frame-freq plane, and a cell having a maximum accumulated value is sequentially selected. To track the trajectory of the peak point. At this time, a cell is selected according to the following conditions (1) to (3).
(1) The frame (time coordinate) is advanced, and if the value at the same frequency (same frequency coordinate) is the maximum value, the cell (frequency) is selected.
(2) If an accumulated value larger than the accumulated value of the same frequency exists on either the left or right side, a cell (frequency) having the largest accumulated value is found and selected.
(3) If the accumulated value of the same value as the accumulated value of the same frequency is on the left or right, or both, the cell is selected without changing the frequency coordinates.
[0026]
For example, in FIG. 5, first, starting from the peak point of [N, 5], “19” of [N, 6] is the maximum value in the sixth frame (time coordinate “6”). 6] cell is selected. In the seventh frame, “12” of [M, 7] and “12” of [N, 7] are both maximum values, but the [N, 7] cell is selected according to the condition (3) above. To do. Further, since [M, 8] is “20” and [N, 8] is “12” in the eighth frame, the [M, 8] cell is selected. Thereafter, by selecting cells in the same manner, the cells indicated by thick frames in FIG. 5 can be sequentially traced as the locus of the peak point.
[0027]
In the above description, a specific example has been described in which peak tracking is started from the peak point [N, 5], but it goes without saying that similar processing is performed for a plurality of peak trajectories. That is, a base point frame serving as a base point for peak tracking is determined, and a track tracking process is performed from each peak point on the base point frame. In the process of determining the base point frame, a time point at which the envelope of the time waveform in FIG. 6 is maximized (hereinafter referred to as an attack max point) is detected. This is because at the attack max point where the amplitude level is maximum, there is a high possibility that all peak points that need to be tracked are available. In the time waveform of FIG. 6, the attack max point can be found at 42.9 ms, and since the sampling frequency is 44.1 kHz, 42.9 ms is obtained as the 1892th point. Therefore, the frame whose center is closest to this attack max point (1892nd point) is detected, and this is set as the 0th frame as the base frame for tracking start of peak tracking. Then, peak tracking is performed from the 0th frame toward the frame later in the future (future), and when the upper limit of the analysis range is reached, the lower limit of the analysis range is reached from the 0th frame toward the previous (past) frame in time. Follow the peak until it reaches. Thereby, it becomes possible to detect without giving a peak that needs to be tracked.
[0028]
When the peak point tracking process is completed as described above, in step S6, based on the track tracking result of FIG. 5, the data is edited into a format suitable for re-synthesis of the sound (sine wave addition synthesis), and the analysis result The data is stored in the data memory 3 and the process is terminated. In this data editing process, for example, the peak point data corresponding to the locus in FIG. 5 is extracted from the peak point data in FIG. , Phase data) is used as analysis result data.
[0029]
As described above, each peak point is arranged on the frame-freq plane, and the amplitude is set with a weight (amplitude coefficient) according to the distance from the peak point in the vicinity of the peak point. Since the amplitude value is accumulated and the peak locus is traced following the extreme value (maximum value in the above example) of the accumulated value, a naturally connected locus can be traced.
[0030]
In the above embodiments, the frame-freq plane has been described as being set in a predetermined area of the data memory 3 as a cell space. However, as such a cell space, for example, a graphic memory or a video memory of a personal computer is used. Etc. may be used to graphically display the peak point of the frame-freq plane, the accumulated value of the amplitude, and the like. In this case, the accumulated value can also be executed by image processing. Further, FIG. 3 showing the frame-freq plane is a conceptual diagram, and these cell spaces may be set in such a format in the storage means. The same applies to the amplitude coefficient template of FIG.
[0031]
In the embodiment, waveform synthesis, musical sound generation, etc. are performed after waveform analysis, but the present invention relates to waveform analysis, and is not limited to the presence or absence of processing such as waveform synthesis or generation of musical sounds. Absent. Further, the residual waveform data need not be generated.
[0032]
Moreover, although the case where the musical tone waveform analysis program is stored in the program memory 2 in advance has been described in the embodiment, the present invention is not limited to this and may be as follows. For example, a musical sound waveform analysis program is recorded on a CD-ROM, and the musical sound waveform analysis program is loaded from a CD-ROM device onto a hard disk. Then, the CPU system 1 develops the musical tone waveform analysis program of the hard disk in a RAM or the like, and controls the operation of the musical sound waveform analysis based on the program of the RAM as in the above embodiment. Thereby, it is possible to cause the CPU to perform the same operation as when the musical sound waveform analysis program is stored in the program memory. In this way, new installation, addition or version upgrade of the musical sound waveform analysis program can be easily performed. Further, a musical sound waveform analysis program may be recorded on a floppy disk, a magnetic disk (MO), etc., and supplied to a RAM or a hard disk.
[0033]
Alternatively, the musical sound waveform analysis program may be downloaded using the network interface 7. In this case, for example, it is connected to a network 9 such as a LAN (local area network), the Internet, or a telephone line, and through the network 9, a musical sound waveform analysis program is received from a server computer and recorded on a hard disk. And the download is complete. Furthermore, the musical sound waveform analysis program may be executed through a network.
[0034]
The present invention is not limited to the personal computer as in the above embodiment, but various devices such as various electronic musical instruments, tone generators, sequencers, and effectors, and these devices are connected to each other using communication means such as MIDI or various networks. Such a system may be incorporated as a function or an application.
[0035]
It is to be noted that a distribution destination such as a medium on which a musical sound waveform analysis program as described in the above embodiments is recorded, that is, ROM, RAM, hard disk, CD-ROM, magneto-optical disk, DVD (digital multipurpose disk), or network server computer In the storage device, the musical sound waveform analysis program according to claim 3 of the present invention is recorded. Computer readable records Corresponds to the medium.
[0036]
【The invention's effect】
As described above, the musical sound waveform analyzer of claim 1 of the present invention, the musical sound waveform analysis method of claim 2, or the claim 3 of the present invention. Computer readable According to the execution of the musical sound waveform analysis program recorded on the recording medium, the frequency spectrum of the time-series frame is obtained by performing Fourier transform processing on the musical sound waveform to be analyzed, and the peak point of the amplitude data of the analysis data in each frame Cell space with frequency and time coordinates Is defined in the cell space, and the value is reduced according to the distance defined in the cell space between the cell in the front or rear direction with respect to the center cell and the center cell. Using a template in which such an amplitude coefficient is assigned to each cell, the center of the template is superimposed on each of the peak points, and the amplitude coefficient of each cell of the template at that time is multiplied by the amplitude of the peak point, Each multiplication value obtained by the multiplication is accumulated for each cell in the cell space, and the amplitude accumulated in each cell is accumulated in the time axis direction of the frame. Since the locus of the peak point is tracked by selecting the cell following the extreme value, the accumulated value tends to take an extreme value in the cell on the route having the shortest distance between adjacent peak points. Therefore, the locus of the peak point can be tracked with high accuracy.
[Brief description of the drawings]
FIG. 1 is a block diagram of a musical sound waveform analyzer according to an embodiment of the present invention.
FIG. 2 is a flowchart of waveform analysis processing in the embodiment.
FIG. 3 is a conceptual diagram showing a state in which peak points are arranged on a frame-freq plane in the embodiment.
FIG. 4 is a conceptual diagram showing an amplitude coefficient template in the embodiment.
FIG. 5 is a conceptual diagram showing an accumulation result on a frame-freq plane in the embodiment.
FIG. 6 is a diagram illustrating an example of a time waveform of a C4 key sound of a piano according to the embodiment.
FIG. 7 is a diagram illustrating an example of a peak of a frequency spectrum according to the embodiment.
FIG. 8 is a diagram for explaining a conventional problem.
[Explanation of symbols]
1 ... CPU, 2 ... program memory, 3 ... data memory, 7 ... musical sound synthesis unit

Claims (3)

The frequency spectrum of the time-series frame is obtained by performing Fourier transform processing on the musical sound waveform to be analyzed, the peak point of the amplitude data of the analysis data in each frame is detected, and the peak point continuous in the time axis direction of the frame In a musical sound waveform analyzer that tracks the trajectory of
Storage means in which a cell space defined by frequency coordinates and time coordinates is defined, and at least for time coordinates, a cell space defined as coordinates in units of time intervals dividing the frame into a plurality of equal parts ,
An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is temporally forward or backward with respect to the central cell and the central cell. A calculation unit that uses a template in which a coefficient is assigned to each cell, superimposes the center of the template on each of the peak points, and multiplies the amplitude coefficient of each cell of the template at that time by the amplitude of the peak point;
Accumulating means for accumulating each multiplication value obtained by the computing means for each cell in the cell space;
Peak tracking means for tracking the locus of the peak point by selecting the cell following the extreme value of the amplitude accumulated in each cell in the time axis direction of the frame. Musical sound waveform analyzer. The frequency spectrum of the time-series frame is obtained by performing Fourier transform processing on the musical sound waveform to be analyzed, the peak point of the amplitude data of the analysis data in each frame is detected, and the peak point continuous in the time axis direction of the frame In the musical sound waveform analysis method to track the trajectory of
A cell space composed of cells defined by frequency coordinates and time coordinates, and at least for time coordinates, a cell space defined as coordinates in units of time intervals dividing the frame into a plurality of equal parts is used.
An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is forward or backward in time with respect to the center cell and the center cell. Using a template in which a coefficient is assigned to each cell, the center of the template is superimposed on each of the peak points, and the amplitude coefficient of each cell of the template at that time is multiplied by the amplitude of the peak point,
Accumulating each multiplication value obtained by the multiplication for each cell in the cell space;
Musical sound waveform analysis characterized in that the locus of the peak point is traced by selecting the cell following the extreme value of the amplitude accumulated in each cell in the time axis direction of the frame. Method. The frequency spectrum of the time-series frame is obtained by performing Fourier transform processing on the musical tone waveform to be analyzed, the peak point of the amplitude data of the analysis data in each frame is detected, and the peak point continuous in the time axis direction of the frame A computer-readable recording medium on which a musical sound waveform analysis program for executing a process for tracking a trajectory of a computer is recorded,
A cell space composed of cells defined by frequency coordinates and time coordinates, and at least for time coordinates, the detected cell space is defined using coordinates defined in units of time intervals that divide the frame into a plurality of equal parts . Associating a peak point with a cell in the cell space according to the frequency and frame of the peak point;
The amplitude set for each cell is accumulated for each cell in the vicinity of the peak point by setting the amplitude of the peak point with a weight according to the distance defined in the cell space from the peak point. Steps,
An amplitude that is defined in the cell space and has a value that decreases according to the distance defined in the cell space between a cell that is temporally forward or backward with respect to the central cell and the central cell. a step of coefficients using the template assigned to each cell, the center of the template superimposed on each of the peak point, multiplying the amplitude of the amplitude coefficient and the peak point of each cell in the template at that time,
Accumulating each multiplication value obtained by the multiplication for each cell in the cell space;
Tracing the locus of the peak point by selecting the cell following the extreme value of the amplitude accumulated in each cell in the time axis direction of the frame;
A computer-readable recording medium having recorded thereon a musical sound waveform analysis program for executing on a computer .

Images