JP4294179B2 - Waveform playback device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a waveform reproducing apparatus that performs compression and expansion on a time axis using a so-called phase vocoder.
[0002]
[Prior art]
In general, in a phase vocoder, an audio waveform signal is divided into a plurality of frequency bands having an approximate fundamental period of the signal as a bandwidth, and the signal of each band is multiplied by a complex frequency at the center of each band to obtain an amplitude value. And the instantaneous frequency are stored in the memory. For each band, a sine or cosine oscillation wave oscillated at the sum of the analyzed instantaneous frequency and the center frequency of the corresponding band is amplitude-modulated with the analyzed amplitude value. Thus, the original audio waveform signal is reproduced by mixing all signals generated for each band.
[0003]
It has been proposed to compress and expand the time axis of the original audio waveform signal using such a phase vocoder. For example, by compressing or expanding the amplitude value information and the instantaneous frequency information on the time axis, and recombining the original audio waveform signal using the compressed or expanded amplitude value information and the instantaneous frequency information, It has been proposed to compress or expand the original audio waveform signal on the time axis.
[0004]
FIG. 15 shows a block diagram of an example of a known phase vocoder. FIG. 16 shows a detailed configuration of an analysis unit (band k analysis unit) related to band k (in the example of FIG. 16, k is an integer from 0 to 99) in the phase vocoder of FIG. . Hereinafter, the phase vocoder will be described in detail with reference to FIGS. 15 and 16.
[0005]
The phase vocoder divides an audio signal, that is, a waveform into a plurality of frequency bands (bands) having a bandwidth of the approximate fundamental frequency band of the audio signal (in the vocoder of FIG. 15, the band is shown in FIG. 17). 0 to 100 frequency bands of band 99). The divided audio signal of each frequency band is supplied to the analysis unit of each frequency band, multiplied by the complex frequency at the center of each divided frequency band, and analyzed and developed into an amplitude value and an instantaneous frequency.
[0006]
In the analysis unit shown in FIG. 16, w (n) is an impulse response of the analysis filter. The operation of each analysis unit is equivalent to a short-section Fourier transform cut out by a known w (n) window.
[0007]
The amplitude value and the instantaneous frequency obtained by each analysis unit are stored in the memory.
[0008]
The audio signals in each frequency band stored in the memory are synthesized in a synthesis unit (time frequency conversion processing unit, adder and cosine oscillator shown in FIG. 15). Each synthesis unit modulates the sine wave of the center frequency of each analyzed frequency band according to the analyzed amplitude value and instantaneous frequency, generates an audio signal of each frequency band, When the audio signals are mixed, the original audio signal can be restored.
[0009]
Here, when compressing and expanding the reproduction time of the audio signal, the time-frequency conversion processing unit obtains the interpolation value of the amplitude value and the interpolation value of the instantaneous frequency in the time-frequency conversion processing unit.
[0010]
FIG. 18A shows the configuration of a time-frequency conversion processing unit for band k for performing time-frequency conversion processing for band k. A case where the reproduction time of the audio signal is compressed and expanded will be described with reference to FIG.
[0011]
First, when extending the reproduction time of an audio signal, the time-frequency conversion processing unit interpolates the amplitude value at each sample point, and extends the envelope of the amplitude value based on the time expansion / contraction information. Also, the interpolated value of the sample point is obtained based on the time expansion / contraction information (see FIG. 18B). From the amplitude value and the instantaneous frequency obtained by the interpolation in this way, the audio signals of the divided frequency bands are obtained and mixed in the same manner as described above.
[0012]
On the other hand, when the reproduction time of the audio signal is compressed, the amplitude value and the instantaneous frequency are thinned out by interpolation based on the time expansion / contraction information, and the envelope is reduced (see FIG. 18C). From the amplitude value and the instantaneous frequency obtained by the interpolation in this way, the audio signals of the divided frequency bands are obtained and mixed in the same manner as described above.
[0013]
When the pitch of the audio signal is changed, the sum of the center frequency and the instantaneous frequency of each divided frequency band is multiplied by the frequency conversion information (change ratio), and the above-described interpolation calculation is executed.
[0014]
Since the above-described processing is executed by a known method, a flowchart and a small explanation thereof are omitted.
[0015]
In such compression and expansion, the entire audio waveform signal is uniformly compressed or expanded on the time axis. Therefore, if the original audio waveform signal is, for example, a musical tone signal, the attack portion thereof is also compressed and expanded. In this case, the musical tone may be very unnatural.
[0016]
In order to prevent this, when compressing the time axis by performing waveform reproduction based on the amplitude value information and instantaneous frequency information as usual, without compressing and expanding the amplitude value information and instantaneous frequency information itself of each band. When the audio waveform signal is speech, when the set compression time has elapsed for a certain syllable, the process proceeds to waveform resynthesis of the next syllable, and when it is expanded, the amplitude for each band The repetition period between the value information and the instantaneous frequency information is determined in advance, and after reading out the repetition period of a syllable between the amplitude value information and the instantaneous frequency information, the amplitude value and the instantaneous frequency information of the repetition period are obtained. It is conceivable to re-synthesize a waveform by repeatedly using it, and re-synthesize a waveform of a certain syllable until a set expansion time elapses.
[0017]
[Problems to be solved by the invention]
However, when such repetitive playback is performed, the period of the repetition period in each band and the position of the repetition period do not always match, and the timing to advance to the next syllable does not match in each band, Become unnatural. Also, due to the configuration of the phase vocoder that uses instantaneous frequency information, even if an attempt is made to re-synthesize the original audio waveform signal without compression or expansion, the original audio signal is re-synthesized because it does not have phase information. I couldn't. In particular, the difference may appear clearly in an audio signal with a sharp rise.
[0018]
In the present invention, when performing time-axis compression / expansion of an audio waveform signal using a phase vocoder and performing reproduction such that an error occurs in the reproduction position of the waveform data of each band, To provide a waveform reproduction device that improves the sound quality of the synthesized waveform by correcting the error in the original, and by re-synthesizing the original audio waveform signal when neither compression nor expansion is performed. Objective.
[0019]
[Means for Solving the Problems]
  The waveform reproduction apparatus according to the present invention includes one or more pieces of mark information indicating a position on the time axis serving as a boundary of the section of the original audio waveform signal including a plurality of sections, and the original audio waveform signal having a plurality of frequencies. Waveform data storage means for storing the phase information and amplitude value information of the waveform signal for each of the divided bands is provided. The phase information and amplitude value information of each band are acquired for each sampling value obtained by sampling the original audio waveform signal at an appropriate sampling frequency. That is, it is acquired corresponding to the position on the time axis of the original audio waveform signal. First reproduction position information generating means for generating first reproduction position information that represents a position on the time axis of the original audio waveform signal, for example, the position of each sampling value, and changes with time at a desired speed. Is provided. This speed is higher when time compression is performed than when the original audio waveform signal is not compressed and expanded on the time axis, and is slower when time is expanded. Second reproduction position information generating means for generating second reproduction position information that represents the position on the time axis of the phase information and amplitude value information of each band and that changes temporally at a speed different from the desired speed. Is provided. The speed of the second reproduction position information can be a constant speed, for example, the speed when the original audio waveform signal is not compressed or expanded on the time axis. Waveform signal synthesizing means is provided for reading the phase information and amplitude value information from the waveform data storage means according to the second reproduction position information, and synthesizing the reproduced audio waveform signal according to the read phase information and amplitude value information. ing. The second reproduction position information is provided independently for each band. Before the position represented by the first reproduction position information reaches the position represented by the mark information in the waveform data storage means, the loop is obtained for the divided band in which the second reproduction position information has reached the end of the loop section. The second reproduction position information is controlled so as to repeatedly read the section, and when the position represented by the first reproduction position information reaches the position represented by the mark information in the waveform data storage means, Control means is provided for controlling the second reproduction position information so as to change to the second reproduction position information corresponding to the represented position.
[0020]
Further, the waveform signal synthesizing means converts the phase information read from the waveform data storage means into frequency information, inputs the converted frequency information, and has a frequency cycle corresponding to the frequency information. A periodic signal generating means for generating a signal and controlling the phase of the periodic signal generated according to the read phase information in response to the change of the second reproduction position information; and the periodic signal generating And amplitude control means for controlling the amplitude of the periodic signal generated from the means in correspondence with the read amplitude information.
[0021]
In the waveform reproduction apparatus of the present invention, the first reproduction position information changes at a faster speed than when the original waveform signal is reproduced as it is when the time axis compression is performed, and when the expansion is performed, It changes at a slow speed. On the other hand, since the second reproduction position information is read out of the amplitude value information and the phase information of each band at a speed different from that of the first reproduction position information, for example, on the time axis specified by the first reproduction position information. And a position on the time axis of the original audio waveform signal corresponding to the amplitude value information and the phase information specified by the second reproduction position information are caused by deviation (deviation). Then, when the first reproduction position information reaches the position represented by the mark information, that is, when a predetermined time axis compression or expansion is performed, the first reproduction position information and the second reproduction position information are Since the control means controls the second reproduction position information so that the deviation becomes zero, the second reproduction position information corresponds to the phase information and amplitude value information of each band corresponding to the mark information. The waveform is corrected to the position and subsequent waveform reproduction is performed based on the phase information and amplitude value information of each band after the position corresponding to the mark information.
[0022]
When the time axis is extended, before the first reproduction position information reaches the mark information position, the second position information indicates the phase information and amplitude value information corresponding to the mark information position. It may be read out. Therefore, a repetition period is determined for the second reproduction position information before the position where the phase information and amplitude value information corresponding to the position of the mark information is stored, and the first reproduction position information is the mark information. Before reaching the position, phase information and amplitude value information in this repetition period can be repeatedly read out. In this case, when the first reproduction position information reaches the position of the mark information, the second reproduction position information is immediately corrected to a position corresponding to the mark information. Also, when performing time axis compression, when the first reproduction position information reaches the position of the mark information, the second reproduction position information stores the storage position of the phase information and amplitude value information corresponding to the position of the mark information. May not be specified. Also in this case, the second reproduction position information is immediately corrected to the storage position of the phase information and the amplitude value information corresponding to the position of the mark information.
[0023]
DETAILED DESCRIPTION OF THE INVENTION
As shown in FIG. 1, the waveform reproducing apparatus according to one embodiment of the present invention has a waveform memory 2 and a DSP 4, which function as a phase vocoder. The phase vocoder synthesizes a reproduced audio waveform signal based on the waveform data stored in the waveform memory 2 and supplies the synthesized audio waveform signal to the D / A converter 5. The DSP 4 operates according to a program stored in the ROM 8. This program is transferred to the DSP 4 via the CPU 6.
[0024]
The CPU 6 detects the operation state of the operation element 10 according to the program stored in the ROM 8, and controls the DSP 4 according to the detection result. The CPU 6 displays the detected operation state of the operation element 10 and the control state of the DSP 4 based on the operation state of the operation element 10 on the display device 12. The operation element 10 is provided with a reproduction mode switch for setting a reproduction mode, a compression / decompression operation element for setting a parameter representing compression or expansion and its degree, a keyboard having a plurality of keys, and the like. The RAM 14 is used as a working memory for the CPU 6, and various registers and flags to be described later are set.
[0025]
Unlike the above-described phase vocoder, the waveform memory 2 stores amplitude value information and phase information. As is apparent from FIG. 11 and FIG. 12 showing the details of each analysis unit shown in FIG. 11, these pieces of information are obtained by sampling each waveform data signal, for example, one phrase of a musical sound sampled at a predetermined sampling frequency. The value x (n) is divided into a plurality of, for example, m frequency bands having an approximate fundamental period of the waveform data signal as a bandwidth, and the signals in each band are multiplied by the complex frequency at the center of each band, Analysis and development of amplitude information and phase information.
[0026]
As shown in FIG. 2, the DSP 4 includes conversion means provided for each band, for example, time frequency conversion processing units 4a-1 to 4a-m, and periodic signal generation means, for example, cosine oscillators 4b-1 to 4b-. m and amplitude control means, for example, function as waveform signal synthesis means comprising multipliers 4c-1 to 4c-m. Further, the DSP 4 also functions as an adder 4d that adds the outputs of the multipliers 4c-1 to 4c-m, and a multiplier 4e that multiplies the added output of the adder 4d with a gate signal. The adder 4d functions as a combiner, and the multiplier 4e functions as a gate.
[0027]
Each of the time-frequency conversion processing units 4a-1 to 4a-m reads amplitude value information and phase information from the waveform memory 2 in accordance with time expansion / contraction information addr of each band supplied from the control means described later, for example, the control unit 4f. The phase information is converted into instantaneous frequency information by the differentiating means, subjected to interpolation calculation, and the corresponding cosine oscillators 4b-1 to 4b-m are oscillated at the sum frequency with the center frequency ωk of the corresponding band. . In addition, when changing a pitch, the conversion ratio which is frequency conversion information is multiplied. Further, in order to set the phase of the cosine oscillators 4b-1 to 4b-m, the phase information read from the waveform memory 2 is supplied to the corresponding cosine oscillators 4b-1 to 4b-m, and the cosine oscillator 4b- As for 1 thru | or 4b-m, a phase is reset by the phase reset signal from the control part 4f to the supplied phase information. In order to implement these processes, the time-frequency conversion processing units 4a-1 to 4a-m can be realized with a configuration as shown in FIG.
[0028]
In FIG. 13, 100 is a reading means for reading amplitude information and phase information from the waveform memory 2 according to the time expansion / contraction information addr, 102 is a differentiating means for differentiating the read phase information, and 104 is from the differentiating means. Interpolation means for interpolating the obtained instantaneous frequency information with time expansion / contraction information addr, which will be described later, 106 is an interpolation means for interpolating the read amplitude value information with time expansion / contraction information addr, and 108 is the interpolated instantaneous frequency information and center frequency An adding means 110 for adding ωk and a multiplying means 110 for multiplying the output from the adding means 108 by frequency conversion information.
[0029]
The cosine oscillators from the cosine oscillators 4b-1 to 4b-m oscillated as described above are converted into the corresponding time frequency conversion processing units 4a-1 to 4a- in the corresponding multipliers 4c-1 to 4c-m. Amplitude modulation is performed according to the amplitude information from m. Thus, the original audio waveform signal can be reproduced by synthesizing the signals generated for each band by the adder 4d. A gate signal for controlling the open / closed state of the gate 4e is supplied from a gate signal generation unit 4g included in the control unit 4f. The gate signal takes a value between 0 and 1, for example, and immediately rises from 0 to 1 when rising, but gradually decreases from 1 toward 0 when falling.
[0030]
The waveform memory 2 stores various parameters including the above-described amplitude value information and phase information. That is, as shown in FIG. 4A, the waveform memory 2 has a waveform information area, a band waveform information area, and a band waveform data area.
[0031]
As shown in FIG. 5B, WaveStart, WaveEnd, Mark (1), Mark (2), Mark (3), and Mark (4) are stored in the waveform information area. Wavestart samples the original audio waveform signal shown in the upper part of FIG. 3 at a predetermined frequency and stores it in the waveform memory 2 (actually not stored in the waveform memory 2). WaveEnd is an address that is also an end address.
[0032]
If Mark (1) to Mark (4) store the sampling values of the original audio waveform signal, for example, in the case of a musical sound, the boundary of each section such as an attack part, a steady part, an attenuation part, and a silent part. In the case of an audio signal, this is the address of the portion that becomes the boundary of each section such as each syllable. These Mark (1) to Mark (4) correspond to mark information. In this embodiment, since the example in which the original audio waveform signal is composed of three sections is shown, there are four pieces of mark information. However, in actuality, the number of pieces of mark information depends on the number of sections of the original audio waveform signal. Increase or decrease.
[0033]
As shown in FIG. 4 (c), the band waveform data area is provided for each of m frequency bands having the approximate basic period of the original audio waveform signal in the bandwidth, as shown in FIG. 4 (d). The amplitude value information and the phase information described above are stored for each band. The amplitude value information and the phase information are stored for each sampling value of the original audio waveform signal, as shown in FIG.
[0034]
The band waveform information area is also provided corresponding to m bands as shown in FIG. 8F, and wave_start, wave_end, mark (1), as shown in FIG. altS (1), altE (1), mark (2), altS (2), altE (2), mark (3), altS (3), altE (3), and mark (4) are stored.
[0035]
wave_start represents a start address among addresses where amplitude value information and frequency information in each band of each sampling value are stored, and wave_end represents an end address. The mark (1) to mark (4) correspond to the boundaries Mark (1) to Mark (4) of each section described above, and amplitude value information and frequency information of sampling values at these positions are stored. Represents an address.
[0036]
altS (1) to altS (3) represent addresses where amplitude value information and phase information at the start point of the loop period in each section are stored, and altE (1) to altE (3) are loop period information. It represents an address where amplitude value information and phase information at the end point are stored. In the loop period, when the time axis is expanded as will be described later, for example, the amplitude value information and the phase information of the final addresses altE (1) to altE (3) of the loop period are read and the waveform synthesis is performed. This is for performing waveform synthesis based on the amplitude value information and the phase information in the loop period when the expansion is not completed.
[0037]
In this embodiment, since the number of sections is 3, the number of marks is 4, the number of altS is 3, and the number of altE is 3. However, if the number of sections is different from 3, The number also changes depending on the number of sections. Further, as apparent from FIG. 3, the position and length (cycle) of the loop period are different for each band.
[0038]
Next, an outline of the operation of the waveform reproducing apparatus will be described with reference to FIG. In the embodiment of the present invention, amplitude information and phase information are stored in the band waveform data area of the waveform memory 2, but in order to make it easy to understand a specific phenomenon (change in pitch), FIG. In FIG. 3, the phase information is changed to frequency information. Actually, the storage information of each band stores phase information P (m) and amplitude information A (m) as shown as a certain band Band (m) in FIG. In this waveform reproducing apparatus, an address counter SPHASE that generates an address (first reproduction position information) of each sampling point of the original audio waveform signal in the DSP 4 and an address (phase information and amplitude value information of each band) are stored. M address counters addr for generating (second reproduction position information) are used.
[0039]
The value of SPHASE increases from WaveStart. The increment tcomp is set according to the degree of compression or expansion set by the compression / decompression operation element included in the operation element 10. For example, in the case of compression, the increment tcomp is set to a value larger than 1, for example, when compression / expansion is not performed, and is set to a value smaller than the increment 1 when compression / expansion is not performed. The Further, each addr increases from mark (1), but the increment rate is 1, which is an increment in the case of normal reproduction without compression / decompression, for example, unlike the SPHASE increment tcomp. Note that the increment rate may be arbitrarily set by an operator.
[0040]
According to the value of each addr, phase information and amplitude value information of each band is read from the waveform memory 2 and supplied to the corresponding frequency time conversion processing units 4a-1 to 4a-m to recombine the waveforms.
[0041]
When SPHASE and each addr generate an address at the same period as the sampling period of the original audio waveform signal, for example, when tcomp is greater than 1, that is, in the case of time compression, SPHASE sets Mark (2). When specified, each addr has not yet reached the position of mark (2) corresponding to Mark (2). At this time, a deviation occurs between the values of each addr and SPHASE. The value of each addr is forcibly corrected to mark (2) so as to make this deviation 0, and at the same time, a phase reset signal is supplied from the control unit 4f to each cosine oscillator 4b-1 to 4b-m. The respective phase information in each mark (2) is set in each cosine oscillator 4b-1 to 4b-m. As a result, the audio waveform signal time-compressed based on the phase information and amplitude value information of each band corresponding to the section starting from Mark (2) is re-synthesized, and the re-synthesized sound becomes unnatural. Absent. Hereinafter, similarly, the time-compressed waveform is recombined in the section after Mark (2).
[0042]
Similarly, when tcomp is smaller than 1, that is, when the time axis is extended, even if each addr reaches mark (2), SPHASE does not reach Mark (2). Therefore, a loop period is provided for each band, and when the value of each addr reaches the address alt_E at the end of each loop period (the end point of the loop period of each band is collectively referred to as alt_E), To alt_S (the beginning of the loop period of each band is collectively referred to as alt_S), phase information and amplitude value information are read in the reverse direction, and the waveform is reproduced. If SPHASE does not reach Mark (2) even after returning to alt_S, the phase information and amplitude value information are read in the forward direction toward alt_E, and the waveforms are recombined.
[0043]
When SPHASE reaches Mark (2) during reading in such a loop period, there is also a deviation between the SPHASE value and each addr value. Therefore, the value of each addr is forcibly corrected to mark (2) corresponding to Mark (2) so that the deviation becomes 0, and at the same time, the phase reset is performed on each cosine oscillator 4b-1 to 4b-m. A signal is supplied to set each phase information in each mark (2) to each cosine oscillator 4b-1 to 4b-m. As a result, the recombination of the time-extended waveform in the section starting from Mark (2) is started based on the phase information and amplitude value information of mark (2) of each band corresponding to Mark (2). Therefore, the start timing of the section starting from Mark (2) is the same for each band, and the re-synthesized sound does not become unnatural. The time axis is similarly extended in the section after Mark (2).
[0044]
In the case of such compression / decompression, since the increment rate of each addr is 1, re-synthesis of the audio waveform signal is performed by reading each sampling value of the audio waveform signal at the sampling frequency. Therefore, for example, when the original audio waveform signal is a musical sound signal, the attack portion is not compressed and expanded, and can be re-synthesized without impairing the sound quality of the attack portion of the original waveform signal, resulting in an unnatural musical sound. Absent.
[0045]
Hereinafter, operations of the DSP 4 and the CPU 6 will be described in detail. FIG. 5 shows the main routine of the CPU 6. The CPU 6 mainly detects the state of the operation element 10, performs various settings according to the detected state, and displays the detection state and the setting state on the display device 12.
[0046]
First, it is determined whether or not the playback mode switch included in the operation element 10 has changed (step S2). There are mode 1 and mode 2 as playback modes. Mode 2 is a mode in which the timing of the frequency information and amplitude value information of each band is matched when re-synthesis of a new section of the original audio waveform signal as described above is started. Mode 1 is such a timing. This mode does not correct.
[0047]
If it is determined that there is a change in the playback mode switch, it is determined whether the changed playback mode switch is mode 1 or mode 2 (step S4). If the playback mode switch is mode 1, the playback mode is set to 1 (step S6). If the playback mode switch is mode 2, the playback mode is set to 2 (step S8). Is changed to the set mode (step S10).
[0048]
Subsequent to step S10 or when it is determined in step S2 that there is no change in the playback mode switch, it is determined whether the compression / decompression operator included in the operator 10 has changed (step S12). If it is determined that there is a change in the compression / decompression operator, the compression / decompression parameter value tcomp is set corresponding to the operation amount of the compression / decompression operator (step S14). Next, the display of the value of tcomp on the display device 12 is updated (step S16).
[0049]
Subsequent to step S16 or when it is determined in step S12 that there is no change in the compression / decompression operator, it is determined whether or not the key operator of the keyboard included in the operator 10 has been operated (step S18). . If it is determined that the key operator has changed, whether the change of the key operator is off to on (sounding starts), on to off (sounding stops), or on to on (legato performance). Is determined (step S20). If it is determined that the change is from off to on, the frequency conversion information is calculated from the note number that is the MIDI data corresponding to the pressed key, and the frequency conversion information is transferred to the DSP 4 (step S21). Next, an instruction is sent to the DSP 4 to execute the DSP sound generation start process (step S22). If it is determined in step S20 that the change is from on to on, it means that the key has been played legato, and new frequency conversion information is calculated from the note number which is MIDI data corresponding to the new key press. Then, the frequency conversion information is transferred to the DSP 4 (step S23). Similarly, if it is determined that the change is from on to off, an instruction is sent to the DSP 4 to execute DSP sound generation stop processing (step S24). This completes the main routine. The main routine is repeatedly executed every predetermined period.
[0050]
In the DSP sound generation start process, initial setting is performed (step S26). That is, as shown in FIG. 6, the value of the address counter SPHASE that specifies the address of each sample value of the original audio waveform signal is set in the WaveStart that is the start address. Next, the value of the register mk that stores the mark information at the head of the section to which the address currently represented by SPHASE belongs is set to Mark (1). Then, after the initial setting, the value of the register mk_old that stores the mark information indicating the head of the section immediately before the section to which the address currently generated by the address counter SPHASE belongs is Mark (1). Next, the value of the counter n for designating each mark information is set to 1. A flag key_on indicating whether or not sound generation has started is set to 1 to indicate that sound generation has started. Then, a flag dir for designating a direction for reading the amplitude value information and frequency information in the loop period is set in the forward direction.
[0051]
Then, the DSP2 interrupt is permitted (step S28). This completes the DSP sound generation start processing routine.
[0052]
As shown in FIG. 7, the DSP sound generation stop processing routine sets key_on to 0 in order to stop sound generation (step S30). Next, the gate signal generator 4f is instructed to fall the gate signal (step S32). As a result, the output of the gate 4e is gradually reduced and muted.
[0053]
The DSP interrupt processing routine is executed each time a clock signal having a frequency equal to the sampling frequency of the original audio waveform signal is generated by a clock generator (not shown). In this routine, as shown in FIG. 8, first, a control main process is executed (step S34). The control main process will be described later.
[0054]
Next, the value of the counter j for designating each band is set to 1 (step S36), and the time axis conversion processing of the band designated by the counter j is performed as described later (step S38). This time axis conversion process will also be described later. It is determined whether the value of the counter j is greater than m (step S40). If it is greater than m, the interrupt process is terminated. If it is not greater than m, the value of the counter j is incremented by one and step S38 is executed. Then, it is determined whether a flag syl_flg described later is 1 (step S39). When the flag syll_flg is 1, the phase reset signal is supplied to the cosine oscillators 4a-1 to 4a-m to instruct to reset the phase, and the interrupt processing routine of the DSP 4 is ended (step S41). If it is determined in step S39 that the flag syl_flg is not 1, the interrupt processing routine of the DSP 4 is terminated.
[0055]
The DSP control main process is mainly for advancing the value of SPHASE by tcomp each time a clock signal is generated. As shown in FIG. 9, in the DSP control main process, a flag that is set to 1 only when the address counter SPHASE reaches any one of Mark (1) to (4), that is, when it reaches the head of each section. It is first determined whether syl_flag is other than 0 (step S42). If it is other than 0, this is set to 0 (step S44). Following this, or when it is determined in step S42 that syl_flag is not 0, that is, 0, it is determined whether the value of the address counter SPHASE is greater than or equal to mk (step S46). Initially, in the DSP sound generation start processing, SPHASE is set to WaveStart and mk is set to Mark (1). Therefore, when step S46 is executed for the first time, it is determined that the value of the address counter SPHASE is greater than or equal to mk. .
[0056]
If the value of the address counter SPHASE is greater than or equal to mk, it is determined whether SPHASE is greater than or equal to WaveEnd, that is, whether the last address of the section has been designated. If the final address is not specified, syl_flag is set to 1, and the value of the counter n is incremented by 1 (step S50). That is, SPHASE is located at the head of a certain section.
[0057]
Next, mk is updated to Mark (n) (step S52). Thereby, the value of mk represents the head of the next section. The gate signal generator 4g is instructed to rise the gate signal (step S54). As a result, a reproduction waveform can be output from the gate 4e.
[0058]
Then, it is determined whether key_on is 0 (step S56). The key_on becomes 0 only when the DSP sound generation stop process described above is executed. If key_on is not 0, the value of the address counter SPHASE is increased by tcomp (step S58), and this control main process is terminated.
[0059]
If key_on is 0, it is determined whether the gate signal is 0 (step S60). Since the gate signal is gradually lowered after being instructed to fall, it is determined whether the gate signal is actually 0 and no output is generated from the gate 4e. If it is determined that the gate signal is not 0, step S58 is executed. If the gate signal is 0, the DSP interrupt process is stopped (step S62), and the process returns. Therefore, the interrupt process is not executed even if the clock signal is generated until the interrupt is permitted again.
[0060]
If it is determined in step S48 that SPHASE is equal to or greater than WaveEnd, step S62 is immediately executed.
[0061]
If it is determined in step S46 that SPHASE is not mk or more, that is, if it is determined that the head of the next section has not been reached, it is determined whether the value of SPHASE is mk−α or more. That is, it is determined whether the SPHASE has reached the previous address by a predetermined address distance α from the head of the next section (step S64). α is used to determine the position where the reproduction synthesized sound starts to be muted immediately before the end of each section. If it is determined that the SPHASE value is not greater than or equal to mk-α, steps S56, S60, S58 or S56, S60, S62 are executed.
[0062]
If the SPHASE value is greater than or equal to mk−α, it is determined whether the value of mk is not equal to mk_old (step S66). When SPHASE designates the head address of a certain section, mk is updated to the head address of the next section by step S52. Therefore, when step S66 is executed for the first time, mk does not match mk_old. Accordingly, the gate signal generation unit 4g is instructed to fall the gate signal (step S68). Then, mk_old is updated to the value of mk (step S70), and step S56 and subsequent steps are executed. Accordingly, when step S66 is executed after the second time, step S56 and subsequent steps are executed immediately after step S66.
[0063]
In this manner, in the DSP control main routine, SPHASE is incremented by tcomp every time a clock signal is generated, and only when SPHASE designates the head address of each section, syl_flag is set to 1, enabling output from the gate 4e, and SPHASE. When the address of mk−α is designated for the first time, the gate signal starts falling.
[0064]
As shown in FIG. 10, in the DSP time axis conversion process, it is determined whether the set reproduction mode is 1 or 2 (step S72). In the reproduction mode 1, the addr of the band designated by the j counter is increased by tcomp (step S74), and the addr is transferred as time expansion / contraction information to each time frequency conversion processing unit (step S76).
[0065]
If the playback mode 2 is set, it is determined whether syl_flag is 1, that is, whether the head address of any section is specified (step S78). If the head address of any section is specified, addr is updated to mark (n−1) in order to set the head address of the section to addr (step S80).
[0066]
This is because, when syll_flag is set to 1 in step S50 of the DSP control main process (when SPHASE reaches the beginning of any section of the original audio waveform signal), n is advanced by 1, so addr is marked ( If n), the address at which the amplitude value information and the phase information at the head of the section next to the section specified by SPHASE are stored is designated. Therefore, an address where amplitude value information and phase information are stored at the head of the section of the original audio waveform signal designated by SPHASE is designated as mark (n−1).
[0067]
Thus, when SPHASE designates the head of a certain section, addr is also forcibly corrected to the address where the amplitude value information and phase information of the head of the section are stored. Regardless of whether the decompression is performed, the reproduction synthesis of the new section is started by the phase information and the amplitude value information of each band of the head sampling value of the section.
[0068]
Next, the alt_start value for storing the start address of the loop period is designated as altS (n-1) of the band designated by the j counter, and the alt_end value for storing the address at the end of the loop period is designated by the j counter. The bandwidth is updated to altE (n-1) (step S82). This is also for storing the beginning and end of the loop period of the section to which the address specified by addr belongs. Next, dir is set in the forward direction (step S84), and step S76 is executed.
[0069]
If it is determined in step S78 that syll_flag is not 1, SPHASE does not designate the head of the section, so it is determined whether dir is in the forward direction (step S86). In the forward direction, addr is increased by a predetermined rate (step S88). As the rate, for example, 1 is used as described above. Therefore, according to the addr transferred in step S76, the amplitude information and the phase information are read out by the time-frequency conversion processing unit at a normal speed.
[0070]
It is determined whether addr is greater than or equal to alt_end (step S90). If not greater than alt_end, step S76 is executed. If addr is greater than or equal to alt_end, an address that exceeds the end of the loop period is specified, so the value of addr is subtracted from alt_end by the difference between the current addr value and alt_end (addr-alt_end). (Step S92). Next, dir is set in the reverse direction (step S94), and step S76 is executed.
[0071]
The calculation of step S92 changes the read direction of addr in the reverse direction in step S94. At this time, in the forward direction, by the amount the addr has advanced beyond alt_end, in the reverse direction from the address returned from alle_end, This is for reading the frequency information and the amplitude value information.
[0072]
If it is determined in step S86 that the direction is opposite, addr is decreased by rate (step S96). It is determined whether addr is less than or equal to alt_start (step S98). If not more than alt_start, step S76 is executed. If addr is less than or equal to alt_start, the address exceeding the beginning of the loop period is specified, so the value of addr is corrected to a value that increases alt_start by the difference between the current addr value and alt_start (addr + alt_start). (Step S100). Next, dir is set in the forward direction (step S102), and step S76 is executed.
[0073]
In step S100, the addr reading direction is changed in the forward direction in step S102, but the frequency information and amplitude value information from the address returned from ale_start in the forward direction by the amount that the addr has advanced beyond alt_start in the reverse direction. It is for reading out.
[0074]
Since the waveform is reproduced in this way, when the time axis compression / expansion of the original audio waveform signal is performed, when re-synthesizing the head part of each section of the original audio waveform signal, Since the waveform reproduction is performed based on the phase information and the amplitude value information, the sound quality of the synthesized waveform can be improved.
[0075]
In the above embodiment, the playback mode 1 and the playback mode 2 are provided. However, the playback mode 1 is not necessary depending on circumstances. When the reproduction mode 1 is not provided, the waveform memory 2 does not have to store amplitude value information and phase information for the period α in each section.
[0076]
【The invention's effect】
As described above, according to the present invention, when the first reproduction position information reaches the position of the mark information, the second reproduction position information is corrected to the first reproduction position information. The time-axis compression / decompression of the audio waveform signal is performed based on the phase information and amplitude value information corresponding to the mark information, and the time-axis compression / decompression of the audio waveform signal thereafter is similarly performed based on the corresponding phase information and amplitude value information. As it is done, the sound quality is improved.
[Brief description of the drawings]
FIG. 1 is a block diagram of a waveform reproducing device according to a first embodiment of the present invention.
FIG. 2 is a block diagram of a phase vocoder realized by a DSP in the waveform reproduction apparatus of FIG.
3 is a diagram showing frequency information and amplitude value information of each band stored in the waveform memory 2 of the waveform reproducing device of FIG. 1 and an audio waveform signal from which these are generated. FIG.
4 is a diagram showing various data stored in a waveform memory 2. FIG.
FIG. 5 is a flowchart of a main routine executed by the CPU of FIG.
FIG. 6 is a flowchart of a DSP sound generation start processing routine executed by the DSP.
FIG. 7 is a flowchart of a DSP sound generation stop processing routine executed by the DSP.
FIG. 8 is a flowchart of an interrupt processing routine executed by the DSP.
FIG. 9 is a flowchart of a control main process in the interrupt process routine of FIG.
10 is a flowchart of time axis conversion processing in the interrupt processing routine of FIG.
11 is a block diagram of a configuration for storing amplitude value information and phase information in a waveform memory in the waveform reproducing device of FIG. 1; FIG.
12 is a detailed block diagram of an analysis unit in FIG. 11. FIG.
13 is a detailed block diagram of a time-frequency conversion processing unit in the waveform reproduction device of FIG.
14 is a diagram showing phase information and amplitude value information stored as band waveform data in a waveform memory in the waveform reproducing device of FIG. 1; FIG.
FIG. 15 is a block diagram of a conventional phase vocoder.
16 is a detailed block diagram of the analysis unit of FIG.
FIG. 17 is a diagram illustrating waveform analysis performed by the phase vocoder of FIG. 15;
18 is a detailed block diagram of the time-frequency conversion processing unit in FIG. 15 and an explanatory diagram of time expansion and compression in the time-frequency conversion processing unit in FIG.
[Explanation of symbols]
2 Waveform memory (waveform data storage means)
4 DSP (first reproduction position information generation means, second reproduction position information generation means, waveform signal synthesis means, control means).

Claims (2)

One or more pieces of mark information indicating a position on the time axis that is a boundary of the section of the original audio waveform signal composed of a plurality of sections, and each divided band obtained by dividing the original audio waveform signal into a plurality of frequency bands Waveform signal phase information and amplitude value information are stored in correspondence with the position on the time axis of the original audio waveform signal, and the waveform signal for each of the divided bands is a waveform in which different loop sections are set. Data storage means;
First reproduction position information generating means for generating first reproduction position information representing a position on the time axis of the original audio waveform signal and changing in time at a desired speed;
Second reproduction position information that represents the position of the phase information and amplitude value information on the time axis and changes with time at a speed different from the desired speed is generated for each waveform signal in the divided band. Second reproduction position information generating means;
Waveform signal synthesizing means for reading out the phase information and amplitude value information from the waveform data storage means according to second reproduction position information, and synthesizing a reproduction audio waveform signal according to the read phase information and amplitude value information;
Before the position represented by the first reproduction position information reaches the position represented by the mark information in the waveform data storage means, the loop is obtained for the divided band in which the second reproduction position information has reached the end of the loop section. The second reproduction position information is controlled so as to repeatedly read the section, and when the position represented by the first reproduction position information reaches the position represented by the mark information in the waveform data storage means, the mark information Control means for controlling the second reproduction position information so as to change to the second reproduction position information corresponding to the represented position;
A waveform reproducing apparatus provided. 2. The waveform generator according to claim 1, wherein the waveform signal synthesizing means is
Conversion means for converting phase information read from the waveform data storage means into frequency information;
The converted frequency information is input to generate a periodic signal having a frequency corresponding to the frequency information, and the phase of the periodic signal generated according to the read phase information is the second reproduction position information. Periodic signal generating means that is changed and controlled in response to the change of
Amplitude control means for controlling the amplitude of the periodic signal generated from the periodic signal generating means in correspondence with the read amplitude information;
A waveform reproducing apparatus provided.

Images