JP3759810B2 - Method and apparatus for time compression / decompression of waveform data - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a waveform data time compression / decompression method and apparatus for compressing / decompressing the reproduction time of waveform data.
[0002]
One of the editing functions such as sampler is time stretch. This time stretch is a function that compresses or expands a specified range of waveform data to an arbitrary playback time, cuts the original audio waveform into waveform blocks of an arbitrary length, and extends the playback time. Depending on the rate of shortening, part of the waveform block is read twice when decompressing, skipped when compressing, and the previous and subsequent waveform blocks are overlapped and connected, and the overlapped part is crossfade processed Is done.
[0003]
FIG. 9 shows an outline of a method for compressing / decompressing the reproduction time of waveform data. FIG. 9A shows the case where the length L1 of the original speech waveform (hereinafter referred to as the original waveform) is extended (expanded) to the length L2 of the new speech waveform (hereinafter referred to as the playback waveform). In this example, the waveform length L1 is shortened (compressed) to the reproduction waveform length L2.
[0004]
First, the case where the reproduction time of the original waveform is extended will be described with reference to FIG. The original waveform waveform blocks LB1 to LB4 having overlapping portions of LO1 to LO3 (hereinafter collectively referred to as LO) are arranged as shown in the lower stage, and the respective waveform blocks are connected by a crossfade so that the original waveform The reproduction waveform having the length L1 is expanded by time to become L2. The reproduced waveform expanded in time uses the LO portion of the original waveform in an overlapping manner.
[0005]
Next, the case where the reproduction time of the original waveform is compressed will be described with reference to FIG. The waveform blocks LB1 to LB4 of the original waveform excluding the skipped portions of LD1 to LD4 (hereinafter collectively referred to as LD) are arranged as shown in the lower stage, and the respective waveform blocks are connected by crossfading. The reproduced waveform having the length L1 of the original waveform is time-compressed to L2. This time-compressed reproduced waveform is used by deleting the LD portion of the original waveform.
[0006]
[Problems to be solved by the invention]
As described above, the portion where the waveform is discontinuous is usually crossfaded at the joint portion of the waveform block of the playback waveform. When this crossfading is performed, the fading (phase change) of the audio waveform occurs at the crossfade portion. ) Occurs, and this can be heard as an abnormal sound. This abnormal sound is repeated periodically with the period of the length of the playback block, and the period has no correlation with the rhythm of the speech waveform, so it stands out from the reproduced sound and is preferred for the sound quality of the playback waveform for audibility. It has a negative effect.
[0007]
The present invention has been made in view of such problems, and an object of the present invention is to reduce the influence of abnormal sounds generated at the joints of the waveform blocks of the reproduction waveform on the sound quality of the reproduction waveform.
[0008]
[Means for Solving the Problems]
In order to solve the above-described problem, in the time compression / decompression method for waveform data according to the present invention, the waveform data is divided into a plurality of waveform blocks having the same length in time series, and the waveform data is time-compressed. When shortening the length by deleting a part of the block and expanding the time, the part of the waveform block is overlapped to increase the length, and then the waveform data is relocated. A method for time compression / expansion of waveform data for performing time compression / expansion, wherein the length of the waveform block to be divided is determined based on performance timing information (tempo, time signature, and number of beats) of the waveform data. It is a thing.
[0009]
Further, the time compression / expansion device for waveform data according to the present invention has a storage means for storing the waveform data and the length of division based on performance timing information (tempo, time signature, beat number) of the waveform data. Dividing means for deciding and dividing the waveform data into a plurality of waveform blocks of the decided length, and when compressing the waveform data in time, a part of the waveform block is deleted to shorten the length and extend the time In this case, a compression / expansion unit that rearranges the plurality of waveform blocks after increasing the length by overlapping a part of the waveform blocks is provided.
[0010]
In the time compression / decompression apparatus, the performance timing information is stored together with the waveform data in the storage means, and the division means reads the performance timing information from the storage means and determines the length of the waveform block to be divided. Can be configured to determine.
[0011]
[Action]
In the conventional time compression / expansion method, the length for cutting out the waveform block from the original speech waveform is determined regardless of the performance timing (tempo, time signature, number of beats) of the original speech waveform. For this reason, abnormal sounds generated at the joints of waveform blocks in the playback waveform are completely irrelevant to the tempo and time signature of the audio waveform. The sound quality of the sound waveform is impaired in the sense of hearing.
[0012]
Therefore, in the present invention, attention is paid to the length of the waveform block from which the speech waveform is cut out, and the length of the waveform block is determined based on performance timing information of the speech waveform. In this way, the joint of the waveform block can be made regular with respect to the performance timing of the original audio waveform, so even if an abnormal noise occurs at the joint portion of the waveform block of the reproduced waveform, the abnormal sound is Since it correlates with the rhythm of the speech waveform, it is not very noticeable.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram showing a block configuration of a speech waveform processing apparatus equipped with a time compression / decompression apparatus according to the present invention. In this embodiment apparatus, the length for cutting out a speech waveform into a plurality of waveform blocks is determined based on musical timing information (tempo, time signature, number of beats, etc.), and the playback time of the speech waveform is lengthened or shortened. To do.
[0014]
In FIG. 1, a CPU 1 performs processing for extending or shortening the reproduction time of a speech waveform, and processing for the entire apparatus. The ROM 2 stores a control program for the entire apparatus and a control program to be described later for processing to lengthen or shorten the playback time of the audio waveform. The RAM 3 stores voice waveform data and CPU work data. The operation unit 4 and the display unit 5 are used for confirming and inputting parameters for controlling the entire apparatus. The A / D converter 6 and the D / A converter 7 are for inputting and outputting voice waveforms.
[0015]
FIG. 2 shows a configuration example of a panel for operating the embodiment apparatus. In FIG. 2, reference numeral 50 denotes a display screen of the display unit 5. 42 is a rotary encoder used when setting the value of the control parameter, 43 is a button switch for executing or stopping the operation, and 44 is a stretch for instructing a process for extending or shortening the reproduction time of the audio waveform. An instruction switch 45 is a parameter selection switch for selecting the type of control parameter. Reference numeral 46 denotes a waveform selection switch for generating and confirming the sound waveform, and the sound waveform data prepared in advance for each switch in the RAM 3 can be selected.
[0016]
The operation method using these operators will be described. When any one of the waveform selection switches 46 is pressed while the stretch instruction switch 44 is being pressed, the control parameters for the process of lengthening or shortening the reproduction time of the selected audio waveform. Is set, and compression / decompression processing can be executed. The control parameters are displayed on the display screen 50, and the type thereof is switched by operating the parameter selection switch 45. The control parameter displayed on the display screen 50 can be arbitrarily set using the rotary encoder 42. The control parameters that can be set by the rotary encoder 42 are voice waveform data to be assigned to the waveform selection switch 46, a playback range of the voice waveform data (start address S1 and end address E1 of the waveform data), beat number B of the voice waveform data, voice waveform There is a companding ratio R for increasing or shortening the reproduction time, and the number Q of waveform blocks cut out per beat.
[0017]
Hereinafter, the operation of the apparatus according to the present embodiment will be described with reference to the flowcharts of FIGS. FIG. 3 is a flowchart showing a flow of panel processing for extending or shortening the reproduction time of the audio waveform. This panel processing routine is a routine for monitoring the state of the operator on the panel in the main routine of the embodiment apparatus, and is called when one of the waveform selection switches 46 is pressed while the stretch instruction switch 44 is pressed. Executed.
In the following description, the original sound waveform that lengthens or shortens the playback time is referred to as an original waveform, and the sound waveform that is created by lengthening or shortening the playback time is referred to as a playback waveform.
[0018]
First, control parameters are set (step S1). In the control parameter setting, companding is performed to increase or decrease the number of beats B (unit: beat) and playback time with respect to the original waveform assigned to the waveform selection switch 46 pressed together with the stretch instruction switch 44. The ratio R (unit:%) sets the number Q (unit: pieces / beat) of the waveform blocks to be cut out per beat using the control.
Here, the tempo Temp may be set instead of the beat number B of the original waveform, or the tempo Temp of the reproduction waveform may be set instead of the companding ratio R.
[0019]
When the control parameter setting process is completed, a guidance message “Do you want to execute the process?” Is displayed on the display screen 50 (step S2). When “YES” is selected with the button switch 43, the compression / decompression process of step S3 is executed, and when “NO” is selected, the process returns to the main routine.
[0020]
In the audio waveform compression / expansion processing in step S3, the original waveform is divided into a plurality of waveform blocks, the divided waveform blocks are rearranged, and cross-fade processing is performed between the waveform blocks to increase or decrease the reproduction time. Create a playback waveform. Details of the compression / decompression processing will be described later.
[0021]
When the compression / decompression process is completed, a guidance sentence “Do you want to assign a new speech waveform?” Is displayed on the display screen 50 asking whether to save the playback waveform (step S4). When “YES” is selected with the button switch 43, a reproduction waveform is assigned to the waveform selection switch 46 to which the original waveform has been assigned (step S5). If “NO” is selected, the reproduced waveform is not assigned to the waveform selection switch 46, and the process is terminated and the process returns to the main routine.
[0022]
Next, the audio waveform compression / decompression process will be described with reference to the flowchart of the compression / decompression process routine of FIG. 4 and the time charts of the compression / decompression process of FIGS. Here, FIG. 5 shows a method of time expansion by dividing an original waveform into a plurality of blocks and rearranging / crossfading each waveform block, and FIG. 6 shows a method of time compression. In the compression / decompression processing routine of FIG. 4, a plurality of original waveforms are converted using the control parameters set in the panel processing routine of FIG. 3 and the reproduction range of the original waveform data (the start address S1 and end address E1 of the waveform data). The waveform blocks are divided, the divided waveform blocks are rearranged in time series, cross-fading processing is performed between the waveform blocks, and the reproduction time of the original waveform is lengthened or shortened.
[0023]
First, the waveform block length is determined (step S31). In the determination process of the waveform block length, the number N of waveform blocks, the length L1 of the original waveform, the length L1B of the waveform block of the original waveform, the length L2 of the reproduced waveform, and the length L2B of the waveform block of the reproduced waveform are 5 Automatically determine the value of one parameter. Hereinafter, how to make these determinations will be described.
[0024]
The number N (unit: pieces) for dividing the original waveform into a plurality of waveform blocks is calculated using the number of beats B (unit: beats) and the number Q of waveform blocks cut out per beat (unit: pieces / beat).
N = Q × B
Is obtained by calculating. Thereby, the musical timing information of the speech waveform is reflected in the division number N of the waveform block.
[0025]
The length L1 (unit: sample) of the original waveform is determined by using the end address E1 and start address S1 of the waveform data.
L1 = E1-S1
Is obtained by calculating.
[0026]
The length L1B (unit: sample) of the waveform block when cutting the original waveform into a plurality of waveform blocks is:
L1B = L1 / N
Is obtained by calculating. The length L1B of the waveform block corresponds to a “note” having an arbitrary length in the original waveform. For example, if the length of a quarter note is 1 beat, the length L1B of the waveform block is 1 / Q of the length of a quarter note of the original waveform, that is, a length of an integer.
[0027]
The length L2 of the reproduced waveform is obtained by using the length L1 of the original waveform and the companding ratio R of the reproduction time.
L2 = L1 × (R / 100)
Is obtained by calculating.
[0028]
Similar to the original waveform, the length L2B of the waveform block when dividing the reproduced waveform into a plurality of waveform blocks is:
L2B = L2 / N
Is obtained by calculating. The length L2B of the waveform block corresponds to a “note” having an arbitrary length in the reproduced waveform. For example, if the length of a quarter note is 1 beat, the length L2B of the waveform block is 1 / Q of the length of the quarter note of the original waveform, that is, a length of an integer.
[0029]
When the waveform block determination process is completed, the working parameters are initialized (step S32). In this initialization process, the three parameters of the counter C and the pointers P1 and P2 are initialized.
[0030]
First, the value of the counter C that processes the waveform block is set to “0”. The counter C is incremented by one every time rearrangement and crossfading processing is performed on one waveform block in the process of generating a reproduction waveform. When all the waveform blocks are processed, the value of the counter C becomes “N”. Become.
[0031]
The start address S1 of the memory area where the original waveform is stored is set to the pointer P1. Further, the start address S2 of the memory area where the reproduction waveform is stored is set to the pointer P2. These pointers P1 and P2 indicate the head address of the waveform block currently being worked on on the speech waveform.
[0032]
Next, the companding ratio R of the voice waveform is checked to determine whether the voice waveform is expanded, compressed, or not compressed / expanded (steps S33 and S35). If R> 100, decompression processing is performed. If R <100, compression processing is performed. Hereinafter, the decompression process and the compression process will be described in detail.
[0033]
In R> 100, that is, in the expansion process, a process for extending the reproduction time of the original waveform is performed (step S34). First, the data of the waveform block from the position of the pointer P1 on the address where the original waveform is stored to the length L1B is read, and this is read from the position of the pointer P2 on the address in the memory area where the reproduction waveform is stored. Copy (write) over L1B.
[0034]
Further, the twice-read portion of the original waveform (the portion of the length LO described in FIG. 9) is read and connected while crossfading to the end of the copied waveform block data. Specifically, a waveform corresponding to a length (L2B−L1B + LF) from the position (L1B × 2−L2B−LF) from the pointer P1 of the original waveform block, that is, the position [P1 + (L1B × 2−L2B−LF)]. The data is read, and the waveform data is copied over the length (L2B-L1B + LF) from the end position of the copied waveform block by the length LF, that is, from the position of (P2 + L1B-LF) to the length (L2B-L1B + LF) At the time of copying, the crossfade portion (length LF) where the waveform data overlaps is subjected to crossfade processing (step S34). The length LF (unit: sample) of the crossfade portion where the waveform data overlaps is determined to be a predetermined constant value according to the system.
[0035]
This decompression process will be described with reference to FIG. 5. A waveform block (length L1B) corresponding to “notes” having an arbitrary length on the original waveform is a waveform corresponding to “notes” having an arbitrary length on the reproduced waveform. A part of the waveform block (length L1B) of the original waveform, that is, a part of the length LO = (L2B−L1B + LF) from the rear end of the waveform block of the original waveform to the front so as to become the block (length L2B). Read twice, and superimpose the front end portion of the twice read portion on the rear end portion of the waveform block of the original waveform by the length LF, and cross-fading the overlap portion (length LF).
[0036]
On the other hand, in the case of R <100, that is, a speech waveform compression process (step S35), a process of shortening the playback time of the speech waveform is performed (step S36). First, the data of the waveform block of length L2B from the position of the pointer P1 on the address where the original waveform is stored is copied over the length L2B from the position of the pointer P2 on the address to the memory area where the reproduction waveform is stored. To do.
[0037]
Further, the waveform block data of length L1B from the pointer P1 on the address where the original waveform is stored is long from the position of (L1B-LF) from the pointer P1, that is, from the position of [P1 + (L1B-LF)]. The waveform data for LF is copied over the crossfade length LF from the position just before the crossfade length LF from the rear end portion of the copied waveform block, that is, the position of (P2 + L2B-LF). The crossfade portion (length LF) where the waveform data overlaps is crossfade processed (step S36).
[0038]
This compression processing will be described with reference to FIG. 6. A waveform block (length L1B) corresponding to an “note” having an arbitrary length on the original waveform is converted into an “note” having an arbitrary length on the reproduced waveform. A part (length L1B-L2B) of the waveform block (length L1B) of the original waveform is skipped so as to become the corresponding waveform block (length L2B), and the waveform block (length L1B) of the original waveform is further skipped. Read out the portion of the crossfade length LF from the rear end to the front, superimpose the read portion on the portion of the length LF on the rear end side of the waveform block (length L2B) of the reproduction waveform, and cross the overlapped portion. Fade. When the length LD of the skipped portion in FIG. 9B is shown in FIG. 6, it corresponds to LD = (L1B−L2B−LF).
[0039]
When the audio waveform compression / decompression process is completed, the working parameters are updated (step S37). That is, the pointer P1 is increased by the length L1B, and the pointer P2 is increased by the length L2B. Thereafter, the counter C is incremented by one. The process is repeated until the value of the counter C becomes “N”, and when the value becomes “N”, the process ends (step S38). As a result, a reproduction waveform in which the reproduction time of the original waveform is lengthened or shortened is created.
[0040]
Various modifications are possible in the practice of the present invention. For example, in the above-described embodiment, in step S1 of the panel processing routine, the number of beats B, the companding ratio R of the reproduction time, and the number Q of waveform blocks to be cut out per beat are all set using the operators. However, the beat number B may be stored together with the waveform amplitude data as waveform data. The number Q of waveform blocks cut out per beat may be included in the calculation as a constant. In this way, the user only needs to set the companding ratio R in order to lengthen or shorten the reproduction time. In this case, step S1 of the panel processing routine is replaced with the flow shown in FIG. That is, the number of beats B is read from the storage means for storing the voice waveform data and set as the number of beats B of the control parameter (step S11), and the ratio (%) input by the operator is set as the companding ratio R of the control parameter. (Step S12).
[0041]
Also, some waveform data stores tempo information (beats / minute). In this case, step S1 of the panel processing routine of FIG. 3 may be replaced with the flow shown in FIG. That is, the tempo information Temp of the waveform data is read from the storage means, and the beat number B is calculated by multiplying the tempo information Temp (beats / minute) by the waveform length (minutes) (step S14). The memory capacity is obtained as the address length from the start address and end address of the waveform data. Since the sampling frequency of the waveform data is known in advance, the address length can be converted from the sampling frequency as time (minutes). Also in this case, the user only needs to set the companding ratio R when the reproduction time is lengthened or shortened (step S15).
[0042]
【The invention's effect】
As described above, according to the present invention, the abnormal sound generated at the joint portion of the waveform block of the reproduced waveform becomes inconspicuous, and an undesirable influence on the sound quality of the reproduced waveform can be reduced.
[Brief description of the drawings]
FIG. 1 is a diagram showing a block configuration of a speech waveform processing apparatus equipped with a time compression / decompression apparatus for waveform data according to the present invention.
FIG. 2 is a diagram illustrating a configuration example of a panel of an example device.
FIG. 3 is a flowchart showing a panel processing routine in a main routine executed by the embodiment apparatus.
FIG. 4 is a flowchart showing a time compression / decompression processing routine in the panel processing routine of the embodiment.
FIG. 5 is a time chart showing a state of time expansion of waveform data in the embodiment.
FIG. 6 is a time chart showing a state of time compression of waveform data in an example.
FIG. 7 is a flowchart of a changed portion in another embodiment of the present invention.
FIG. 8 is a flowchart of a changed portion in still another embodiment of the present invention.
FIG. 9 is a diagram for explaining a conventional time compression / expansion method.
[Explanation of symbols]
1 CPU (Central Processing Unit)
2 ROM (Read Only Memory)
3 RAM (Random Access Memory)
4 Controller 5 Display Unit 6 A / D Converter 7 D / A Converter 42 Rotary Encoder 43 Button Switch 44 Stretch Instruction Switch 45 Parameter Selection Switch 46 Waveform Selection Switch 50 Display Screen of Display

Claims (3)

When the waveform data is divided into a plurality of waveform blocks having the same length in time series, and when the waveform data is time-compressed, a part of the waveform block is deleted to shorten the length, and when the waveform data is expanded, the waveform block a time compression and expansion method of the waveform data to be time compression / expansion of the waveform data by some with overlapping rearranging the plurality of waveform blocks from longer lengths of,
A time compression / decompression method for waveform data, wherein the length of the waveform block to be divided is determined based on performance timing information of the waveform data. Storage means for storing waveform data;
Division means for determining a division length based on performance timing information of the waveform data and dividing the waveform data into a plurality of waveform blocks of the determined length ;
When the waveform data is time-compressed, a part of the waveform block is deleted to shorten the length, and when the time expansion is performed, the part of the waveform block is overlapped to increase the length and then the plurality of waveforms A waveform data temporal compression / expansion device comprising: compression / expansion means for rearranging blocks.   3. The performance timing information is stored together with waveform data in the storage means, and the dividing means reads out the performance timing information from the storage means and determines the length of the waveform block to be divided. Time compression / decompression device for waveform data.

Images