JP5807419B2 - Signal processing device - Google Patents

Description

  The present invention relates to a signal processing apparatus having a function of performing pitch conversion or tempo conversion of an audio signal.

  As this type of signal processing apparatus, there is a delay modulation type signal processing apparatus that performs pitch conversion of an audio signal by combining compression / expansion of a time axis and crossfading. FIG. 12 is a block diagram showing an example of the configuration of this delay modulation type signal processing apparatus. In this signal processing apparatus, the ring buffer 1 is a storage means realized by a RAM or the like, and is used for storing a sample sequence of an audio signal to be processed. The writing unit 2 sequentially receives samples of the audio signal to be processed, and sequentially writes the received audio signal samples to the ring buffer 1 while sequentially changing the write address for the ring buffer 1 in a certain direction. Writing of the sample string to the ring buffer 1 by the writing unit 2 is performed cyclically. That is, the writing unit 2 writes the sample to the ring buffer 1 while sequentially incrementing the write address, for example, and after the writing to the last address of the ring buffer 1 is finished, Return to the address and write the sample.

  The reproducing unit 3 reads out the sample sequence of the audio signal written in the ring buffer 1 in this way from the ring buffer 1 at a reading speed different from the writing speed, and outputs the sample sequence of the audio signal subjected to pitch conversion. It is. The reproduction unit 3 includes a reproduction reference position update unit 31, a read control unit 32, interpolation units 33 and 34, and a signal synthesis unit 35.

  The reproduction reference position update unit 31 is means for updating the reproduction reference position BA so as to follow the write address of the ring buffer 1 with a predetermined address behind. The read control unit 32 is means for generating modulated playback positions MA1 and MA2 based on the playback reference position BA, and generating a read address for the ring buffer 1 based on the modulated playback positions MA1 and MA2. Here, the modulation reproduction positions MA1 and MA2 are generally not integers but real numbers composed of an integer part and a decimal part. Therefore, the read control unit 32 generates read addresses (integers) of a predetermined number of samples before and after the sample corresponding to each of the modulation reproduction positions MA1 and MA2 based on each of the modulation reproduction positions MA1 and MA2. Interpolation units 33 and 34 are caused to execute interpolation processing for calculating samples corresponding to modulation reproduction positions MA1 and MA2 from samples corresponding to addresses.

  FIG. 13A illustrates the relationship between the reproduction reference position BA and the modulation reproduction positions MA1 and MA2 when the pitch conversion ratio r is greater than 1. Here, the pitch conversion ratio r is the ratio of the pitch of the audio signal after pitch conversion to the pitch of the audio signal written to the ring buffer 1. When the pitch conversion ratio r is larger than 1, the reading control unit 32, as shown in FIG. 13A, the second time gradient (specifically, larger than the first time gradient that is the time gradient of the reproduction reference position BA). (A time gradient that is r times the first time gradient), and then jumps by a predetermined value from the current position to the opposite direction (ie, minus direction) across the reproduction reference position BA. The modulation reproduction position MA1 is updated so that is periodically repeated. Further, the read control unit 32 updates the modulation reproduction position MA2 so as to periodically repeat the change and jump in the second time gradient in a state where the phase is shifted by a half cycle with respect to the modulation reproduction position MA1. Specifically, the read control unit 32 causes the modulation reproduction position MA2 to jump at the center timing between the timings at which the jump of the modulation reproduction position MA1 occurs.

  FIG. 13B illustrates the relationship between the reproduction reference position BA and the modulation reproduction positions MA1 and MA2 when the pitch conversion ratio r is smaller than 1 and the pitch of the audio signal to be processed is converted to a pitch lower than the current pitch. doing. Also in this case, the read control unit 32 generates the modulation reproduction positions MA1 and MA2 by basically the same method as in FIG. 13A except that the jump direction is reversed.

  Then, the read control unit 32 gives the read address of each sample corresponding to a predetermined number of sampling points before and after the modulation reproduction position MA1 to the ring buffer 1, reads the sample from the ring buffer 1, and reads the samples. The interpolation unit 33 is caused to execute an interpolation process for obtaining the sample XMA1 corresponding to the modulation reproduction position MA1. Further, the read control unit 32 gives a read address of each sample corresponding to a predetermined number of sampling points before and after the modulation reproduction position MA2 to the ring buffer 1, reads the sample from the ring buffer 1, and reads these samples. The interpolation unit 34 is caused to execute an interpolation process for obtaining the sample XMA2 corresponding to the modulation reproduction position MA2. Then, the signal synthesizer 35 generates a sample Y of the output audio signal by performing a crossfade process using the samples XMA1 and XMA2.

  FIGS. 14A and 14B are diagrams showing the contents of the cross-fade process performed in each case where the pitch conversion ratio r is larger than 1 and smaller than 1. In FIGS. 14A and 14B, the horizontal axis is the reproduction reference position BA. The reproduction reference position BA changes monotonously with time. Therefore, the horizontal axis also serves as the time axis.

  In the cross-fade processing, the samples XMA1 and XMA2 are multiplied by coefficients C1 and C2 that can take values in the range of “0” to “1”, and a sample Y of the output audio signal is generated by adding the multiplication results. The coefficients C1 and C2 may be determined so that the addition result of both is 1 or may be determined so that the sum of the squares of both is 1. As shown in FIGS. 14A and 14B, the coefficient C1 multiplied by the sample XMA1 becomes the minimum value “0” at the timing when the jump of the modulation reproduction position MA1 occurs, and the modulation reproduction position MA1 becomes the reproduction reference position. A sinusoidal periodic waveform having a maximum value “1” at the timing of crossing BA is formed. The coefficient C2 multiplied by the sample XMA2 is the minimum value “0” at the timing when the jump of the modulation reproduction position MA2 occurs, and the maximum value “1” at the timing when the modulation reproduction position MA2 intersects the reproduction reference position BA. A sine wave-like periodic waveform is formed. The crossfading process using the periodic waveform coefficients C1 and C2 whose phases are shifted from each other by a half cycle in this way is because the waveform generated in the column of the sample XMA1 due to the jump of the modulation reproduction position MA1. This is to prevent the discontinuity of the waveform generated in the column of the sample XMA2 from appearing in the output audio signal due to the jump of the continuous and modulated reproduction position MA2.

  According to the above configuration, each sample sequence of the samples XMA1 and XMA2 is reproduced from the interpolation units 33 and 34 with a time gradient of r times the writing speed to the ring buffer 1. Each sample sequence of the samples XMA1 and XMA2 is a sample sequence having a pitch r times that of the original audio signal. Further, since the sample sequence of the output audio signal is obtained by subjecting each sample sequence of the samples XMA1 and XMA2 to crossfading, the sample sequence of the output audio signal also has a pitch r times that of the original audio signal. Sample column.

  In a signal processing apparatus that performs pitch conversion using such a crossfade, when the pitch conversion of an audio signal having an attack section is performed, the attack section appears multiple times in the output audio signal. There was a problem. Hereinafter, this problem will be described with reference to FIG.

  FIG. 15 exemplifies how the reproduction reference position BA and the modulation reproduction positions MA1 and MA2 change with time. In FIG. 15, the horizontal axis is a time axis, and the vertical axis is an axis indicating the address of the ring buffer 1.

  In the example shown in FIG. 15, a sample of the attack portion of the audio signal is stored at the address Ap in the ring buffer 1. Then, the timing t1 when the modulation reproduction position MA2 rises and crosses the address Ap, the timing t2 after which the modulation reproduction position MA1 rises and crosses the address Ap, and the modulation reproduction position MA2 jumps downward and then rises. At three types of timings t3 that cross the address Ap, samples of the attack portion of the audio signal are read from the ring buffer 1. For this reason, the waveform of the attack portion appears three times before and after the output audio signal. Such an output audio signal is emitted as a tremolo sound with an attack that rises several times in succession.

  In Patent Document 1, as a technique for avoiding such inconvenience, a sample corresponding to one of the modulation reproduction positions MA1 or MA2 is not performed during the period in which the waveform of the attack portion is reproduced from the ring buffer 1 and the crossfade processing is not performed. A technique for reproducing from a sample stored in the ring buffer 1 and outputting it as a sample Y of an output audio signal is disclosed.

Japanese Patent No. 3847150

  According to the technique disclosed in Patent Document 1, for example, in the example of FIG. 15 described above, in the process in which the modulation reproduction position MA1 rises and crosses the address Ap where the attack portion is located, the crossfade processing is not performed and the modulation reproduction position MA1 is reached. The corresponding sample is reproduced as the sample Y of the output audio signal. In this case, the waveform of the attack portion appears in the output audio signal only at timing t2 when the modulation reproduction position MA1 crosses the address Ap where the attack portion is located, and the waveform of the attack portion appears in the output audio signal at other timings t1 and t3. Can be prevented. However, the ideal playback timing of the attack portion of the audio signal is the timing ta at which the playback reference position BA becomes the address Ap. Since the attack part of the audio signal is generated asynchronously with the modulation reproduction positions MA1 and MA2, in general, the timing at which the modulation reproduction position MA1 or MA2 crosses the address Ap where the attack part is located (timing t2 in FIG. 15) is Thus, it deviates from this ideal reproduction timing ta. For this reason, the conventional signal processing apparatus has a problem that a deviation in attack timing due to pitch conversion occurs in the output audio signal. Specifically, when pitch-changing sounds such as playing tan, tan, tan, and attacking rhythm sounds at equal intervals, there is a problem that the sounds cannot be heard at equal intervals.

  The signal processing apparatus that performs pitch conversion has been described above, but the same problem occurs in the signal processing apparatus that performs tempo conversion. For example, in an apparatus that performs tempo conversion by reproducing an audio signal at a speed different from that at the time of recording, the pitch of the audio signal is changed in conjunction with the tempo conversion. That is, when the tempo of the audio signal is multiplied by r, the pitch of the audio signal is also multiplied by r. Therefore, for example, when tempo conversion is performed without changing the pitch by the signal processing device described above, the audio signal is written to the ring buffer 1 at a speed r times that at the time of recording, and the reproduction reference position BA is the same as the writing speed. The modulation reproduction positions MA1 and MA2 are generated for changing the speed and reducing the pitch to 1 / r. Also in this case, the attack timing shift accompanying the pitch conversion similar to the above occurs.

  The present invention has been made in view of the circumstances as described above, and an object of the present invention is to provide a signal processing apparatus capable of reducing a shift in attack timing associated with pitch conversion.

  The present invention detects a buffer for storing an audio signal, writing means for sequentially writing an audio signal to the buffer while changing a write address of the buffer, and an attack portion appearing in the audio signal written to the buffer Attack detecting means for generating the output audio signal by reading out the audio signal from the buffer, and the reproducing means follows the change of the write address of the buffer with a predetermined delay. A reproduction reference position updating means for updating the reproduction reference position so that the reproduction reference position is changed at a second time gradient different from the first time gradient, which is the time gradient of the reproduction reference position, and the second time gradient. The first change that changes with the first time gradient as a whole while periodically repeating the change direction and the jump in the opposite direction. A two-channel audio signal corresponding to the first and second modulation reproduction positions is generated by generating a key reproduction position and generating a second modulation reproduction position out of phase with the first modulation reproduction position. Output audio signal generation means for generating the output audio signal using the two audio signals, and the output audio signal generation means comprises: A means for generating the output audio signal by subjecting two audio signals to cross-fading processing, and an audio corresponding to the first modulation / reproduction position prior to the jump of the first modulation / reproduction position The signal is faded out and the audio signal corresponding to the second modulation reproduction position is faded in, and the second modulation reproduction is performed. A signal synthesizing unit that fades out an audio signal corresponding to the second modulation reproduction position and fades in an audio signal corresponding to the first modulation reproduction position before the position jumps; and the attack detection unit; The timing near the timing at which the playback reference position updated by the playback reference position update means becomes the storage position of the attack portion of the audio signal in the buffer in response to detection of the attack portion is set as the target playback timing of the attack portion. One of the first modulation reproduction position and the second modulation reproduction position is set as the modulation reproduction position for attack reproduction, and the orbit of the modulation reproduction position for attack reproduction after the trajectory correction is the reproduction reference at the target reproduction timing. The orbit correction of the attack reproduction modulation reproduction position so as to cross the position orbit. The audio signal corresponding to the modulation reproduction position for attack reproduction is output to the output audio without performing the clof fade process for a predetermined period with the target reproduction timing in between. There is provided a signal processing apparatus comprising trajectory correcting means for outputting as a signal.

  According to this invention, the sample of the attack part is stored in the buffer at the target reproduction timing near the timing when the reproduction reference position updated by the reproduction reference position updating means becomes the storage position of the attack part of the audio signal in the buffer. Generated based on sample. Therefore, it is possible to reduce the deviation of the attack timing accompanying the pitch conversion.

1 is a block diagram illustrating a configuration of a signal processing device according to a first embodiment of the present invention. It is a block diagram which shows the structure of 32 A of read control parts and signal synthetic | combination part 35A in the embodiment. It is a time chart which shows the operation example of the embodiment. It is a figure which illustrates the waveform of a cymbal sound. It is a figure which illustrates the waveform of a piano sound. It is a figure explaining the difference of the attack part detected by various attack detection algorithms. It is a time chart which shows the 1st operation example of the signal processing apparatus which is 3rd Embodiment of this invention. It is a time chart which shows the 2nd operation example of the signal processing apparatus. It is a block diagram which shows the structure of the signal processing apparatus which is 4th Embodiment of this invention. It is a time chart which shows the aspect of the change of the modulation | alteration reproduction | regeneration position in the same signal processing apparatus. It is a time chart which shows the aspect of the change of the modulation | alteration reproduction | regeneration position at the time of the attack detection in the same signal processing apparatus. It is a block diagram which shows the structural example of the conventional signal processing apparatus. It is a time chart which shows operation | movement of the signal processing apparatus. It is a time chart which shows operation | movement of the signal processing apparatus. It is a figure explaining the problem of the duplication reproduction of the attack part which generate | occur | produces in the same signal processing apparatus.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
<First Embodiment>
FIG. 1 is a block diagram showing the configuration of a signal processing apparatus according to the first embodiment of the present invention. This signal processing device is a device that performs pitch conversion of an audio signal, and is used, for example, for key control (pitch conversion) of an audio signal for karaoke performance or pitch conversion accompanying tempo change. In FIG. 1, parts corresponding to those shown in FIG. 12 described above are denoted by common reference numerals, and description thereof is omitted.

  As shown in FIG. 1, the signal processing apparatus according to the present embodiment includes a ring buffer 1, a writing unit 2, a reproducing unit 3A, an attack detecting unit 4, and a control unit 5.

  The attack detection unit 4 is means for detecting an attack unit appearing in the audio signal written to the ring buffer 1. More specifically, the attack detection unit 4 supplies the read address that the writing unit 2 follows with a predetermined value behind the write address generated for the ring buffer 1 to the ring buffer 1, and the read address The audio signal sample corresponding to is read from the ring buffer 1. The attack detection unit 4 outputs an attack detection signal ATK when an attack unit having a section with a large amplitude or a section with a sudden increase in amplitude appears in the sample sequence of the audio signal read from the ring buffer 1.

  In the reproduction unit 3A, the read control unit 32 of the reproduction unit 3 in FIG. 12 is replaced with a read control unit 32A, and the signal synthesis unit 35 is replaced with a signal synthesis unit 35A. In the reproduction unit 3A, the read control unit 32A, the interpolation units 33 and 34, and the signal synthesis unit 35A generate output audio signal samples Y based on the audio signal samples stored in the ring buffer 1. It constitutes signal generation means.

  FIG. 2 is a block diagram showing the configuration of the read control unit 32A and the signal synthesis unit 35A. As illustrated in FIG. 2, the read control unit 32A includes a modulation reproduction position generation unit 321, a read address generation unit 322, and a trajectory correction unit 323. The signal synthesis unit 35A adds the multipliers 351 and 352 that multiply the samples XMA1 and XMA2 output from the interpolation units 33 and 34 by the coefficients C1 and C2, respectively, and the multiplication results of the multipliers 351 and 352. Thus, an adder 353 that generates a sample Y of the output audio signal and a coefficient generator 354 that supplies coefficients C1 and C2 to the multipliers 351 and 352, respectively. The coefficient generation unit 354 in the signal synthesis unit 35A operates under the control of the trajectory correction unit 323 of the read control unit 32A.

  In the read control unit 32A, the modulation / reproduction position generation unit 321 is means for updating the modulation / reproduction positions MA1 and MA2 based on the pitch conversion ratio r and the reproduction reference position BA. The modulation reproduction position generation unit 321 has a second time gradient obtained by multiplying the first time gradient, which is the time gradient of the reproduction reference position BA, by the pitch conversion ratio r during a period in which no attack is detected by the attack detection unit 4. The modulation reproduction position MA1 is updated so as to periodically repeat the change in and the jump in the direction opposite to the change direction. Further, the modulation reproduction position generation unit 321 periodically performs a change in the second time gradient and a jump in a direction opposite to the change direction at a phase that the modulation reproduction position MA2 is shifted from the modulation reproduction position MA1 by a half cycle. As described above, the modulation reproduction position MA2 is updated. The behavior of the modulation reproduction positions MA1 and MA2 within the period in which this attack is not detected is the same as that of the above-described conventional technique (see FIGS. 12 to 14).

  The read address generation unit 322 supplies the read address of each sample at a predetermined number of sampling points before and after each of the modulation reproduction positions MA1 and MA2 to the ring buffer buffer 1, reads these samples from the ring buffer 1, and interpolates 33. And 34 are caused to execute an interpolation process for generating audio signal samples corresponding to the modulation reproduction positions MA1 and MA2.

  The trajectory correcting unit 323 fades out the audio signal corresponding to the modulation reproduction position MA1 and the audio corresponding to the modulation reproduction position MA2 before the modulation reproduction position MA1 jumps during the period in which the attack detection unit 4 does not detect the attack part. The signal synthesizer 35A performs signal fade-in, and before the modulation reproduction position MA2 jumps, the audio signal fade-out corresponding to the modulation reproduction position MA2 and the audio signal fade-in corresponding to the modulation reproduction position MA1 are signaled. The synthesis unit 35A performs the process.

  The trajectory correcting unit 323 determines that the reproduction reference position BA updated by the reproduction reference position updating unit 31 in response to the attack detection unit 4 detecting the attack unit is the storage position of the attack unit of the audio signal in the ring buffer 1. Is the target playback timing of the attack section. The trajectory correcting unit 323 uses one of the modulation playback positions MA1 and MA2 as an attack playback modulation playback position and the other as a non-attack playback modulation playback position, and the attack playback modulation playback position after the trajectory correction at the target playback timing. The trajectory of the modulation reproduction position for attack reproduction is corrected so that the trajectory of the reproduction crosses the trajectory of the reproduction reference position BA, and the signal synthesizing unit 35A performs a clof fade process over a predetermined period with the target reproduction timing interposed therebetween. Is not executed, and an audio signal corresponding to the modulation reproduction position for attack reproduction is output as an output audio signal. The feature of the present embodiment is the processing content of the trajectory correction unit 323 performed in response to the attack detection.

  3A to 3D are time charts showing an operation example of the present embodiment when the pitch conversion ratio r is larger than 1. FIG. As in FIG. 14, the horizontal axis in FIGS. 3A to 3D is the reference reproduction position BA, and this horizontal axis is also the time axis. Here, FIGS. 3A and 3B illustrate the waveforms of the modulation reproduction positions MA1 and MA2 and the waveforms of the coefficients C1 and C2, respectively, when the control according to the attack detection in the present embodiment is not performed. Is. In the present embodiment, as shown in the figure, coefficients C1 and C2 that change sinusoidally are used for the crossfade process, and the square sum of the coefficients C1 and C2 is always 1. FIGS. 3C and 3D respectively illustrate the waveforms of the modulation reproduction positions MA1 and MA2 and the waveforms of the coefficients C1 and C2 when the control according to the attack detection in the present embodiment is performed. .

  When the attack unit is detected by the attack detection unit 4, the trajectory correction unit 323 obtains a timing Tx at which the reproduction reference position BA becomes the storage position of the attack unit, and uses this timing Tx as the target reproduction timing of the attack unit. Then, in the period before the target reproduction timing Tx, the modulation reproduction position jumping at a timing closer to the target reproduction timing Tx out of the modulation reproduction positions MA1 and MA2 is used as the attack reproduction modulation reproduction position, and the other modulation reproduction position is non-attack reproduction. This is the modulation playback position.

  In the example shown in FIG. 3A, the trajectory correcting unit 323 sets the modulation reproduction position MA1 as an attack reproduction modulation reproduction position and the modulation reproduction position MA2 as a non-attack reproduction modulation reproduction position. The trajectory correction unit 323 corrects the trajectory of the attack reproduction modulation reproduction position and the non-attack reproduction modulation reproduction position and controls the coefficients C1 and C2 generated by the coefficient generation unit 354 as follows.

(1) First, the trajectory correcting unit 323 uses the predetermined time T0 immediately after the attack detection to fade out the audio signal corresponding to the attack reproduction modulation reproduction position MA1 and to perform the audio corresponding to the non-attack reproduction modulation reproduction position MA2. The signal synthesizer 35A is caused to execute a signal fade-in. Specifically, as shown in FIG. 3D, the coefficient C1 for multiplying the audio signal sample XMA1 corresponding to the modulation reproduction position MA1 for attack reproduction is reduced to “0” over time T0, The coefficient C2 for multiplying the audio signal sample XMA2 corresponding to the attack reproduction modulation reproduction position MA2 is increased to "1" over time T0.

(2) When the fade-out and the fade-in in (1) are completed, the trajectory correcting unit 323 jumps the attack reproduction modulation reproduction position MA1 from the current position Mst1 to the target position Men1, as shown in FIG. . At that time, as shown in FIG. 3C, the trajectory correcting unit 323 performs the attack at the target reproduction timing Tx when the attack reproduction modulation reproduction position MA1 after the jump rises with the second time gradient. The jump target position Men1 of the attack reproduction modulation reproduction position MA1 is determined so that the orbit of the reproduction modulation reproduction position MA1 crosses the orbit of the reproduction reference position BA.

(3) Next, as shown in FIG. 3D, the trajectory correcting unit 323 takes a predetermined time T1, and fades in the audio signal corresponding to the attack reproduction modulation reproduction position MA1 and the non-attack reproduction modulation. The signal synthesis unit 35A is caused to execute fade-out of the audio signal corresponding to the reproduction position MA2. The predetermined time T1 and the predetermined time T0 described above are determined such that the sum of both is about half of the required time from the generation timing of the attack detection signal ATK to the target reproduction timing Tx.

(4) The orbit of the modulation reproduction position MA1 for attack reproduction is corrected in this way, and the audio signal corresponding to the modulation reproduction position MA2 for non-attack reproduction and the fade-in of the audio signal corresponding to the modulation reproduction position MA1 for attack reproduction , The trajectory correcting unit 323 fixes the coefficient C1 for the audio signal corresponding to the attack reproduction modulation reproduction position MA1 to “1”, and the audio signal corresponding to the non-attack reproduction modulation reproduction position MA2. The coefficient C2 is fixed to “0”.

  As a result, the signal synthesizer 35A does not perform the cross-fading process, and the audio signal sample XMA1 corresponding to the attack reproduction modulation reproduction position MA1 is output as the output audio signal sample Y. During this time, the sample of the attack part in the audio signal is read from the ring buffer 1, and based on this sample, the sample XMA1 of the attack part is generated by the interpolation part 33 and output as the sample Y of the output audio signal.

  Here, the modulation reproduction position MA1 for attack reproduction coincides with the reproduction reference position BA at the target reproduction timing Tx at which the position indicated by the reproduction reference position BA is the storage position of the attack portion of the audio signal in the ring buffer 1. Therefore, if the read address for the ring buffer 1 is generated using the reproduction reference position BA without using the attack reproduction modulation reproduction position MA1, the attack part is reproduced at the same timing as the attack part reproduction timing.

(5) Next, when the predetermined time T2 has elapsed from the target reproduction timing Tx, the trajectory correcting unit 323 changes the non-attack reproduction modulation reproduction position MA2 from the current position Mst2 to the target position, as shown in FIG. Jump to Men2. At this time, the trajectory correcting unit 323 makes the trajectory of the non-attack reproduction modulation reproduction position MA2 after the jump cross the trajectory of the reproduction reference position BA at the timing when the attack reproduction modulation reproduction position MA1 after the trajectory correction jumps. Furthermore, it is better to determine the target position Men2 that is the jump destination of the non-attack reproduction modulation reproduction position MA2.

(6) After jumping the modulation reproduction position MA2 for non-attack reproduction in this way, the trajectory correction unit 323 jumps the modulation reproduction position MA1 for attack reproduction after the trajectory correction, as shown in FIG. By using the required time T3, the audio signal fades out corresponding to the attack reproduction modulation reproduction position MA1 and the audio signal fades in corresponding to the non-attack reproduction modulation reproduction position MA2.
As a result, the playback unit 3A returns to the normal playback operation using the cross-fade process.

  Although not shown, the operation when the modulation reproduction position MA2 is the modulation reproduction position for attack reproduction and the modulation reproduction position MA1 is the modulation reproduction position for non-attack reproduction is the same as described above. The operation when the pitch conversion ratio r is smaller than 1 is the same as the operation when the pitch conversion ratio r described above is larger than 1.

  As described above, according to the present embodiment, at the time of pitch conversion, the attack portion of the audio signal is reproduced at the target reproduction timing Tx where the reproduction reference position BA is the storage position of the attack portion in the ring buffer 1. Thus, it is possible to reduce the deviation of the attack timing accompanying the pitch conversion. Further, according to the present embodiment, since the cross fading is not performed during the reproduction of the attack part, it is possible to prevent the attack part from appearing in the output audio signal. In addition, according to the present embodiment, the crossfade is terminated at least a predetermined time before the target reproduction timing Tx of the attack unit in which the reproduction reference position BA becomes the storage position of the attack unit, and it corresponds to the modulated reproduction position for attack reproduction. Since only the audio signal is output as the output audio signal, and the crossfade is resumed when the predetermined time T2 has elapsed from the target playback timing Tx of the attack section, the attack waveform in which the characteristic of the musical tone appears is reproduced without loss as much as possible. can do.

Second Embodiment
Next, a second embodiment of the present invention will be described. In the first embodiment, the trajectory correcting unit 323 obtains the storage position of the attack unit in the ring buffer 1 based on the timing when the attack detection unit 4 detects the attack unit, and the reproduction reference position BA is the storage position of the attack unit. The target reproduction timing Tx of the attack portion is set as the timing of the above, and the orbit correction of the attack reproduction modulation reproduction position is performed based on the target reproduction timing Tx.

  However, the musical sound waveform to be subjected to pitch conversion includes, for example, a short attack period occupied by the attack part, such as a cymbal sound shown in FIG. 4, or an attack occupied by the attack part, such as a piano sound shown in FIG. There is a long period. There are various attack detection algorithms. For example, as shown in FIG. 6, an algorithm for detecting the timing A1 immediately after the start of the attack when the sound pressure is slightly increased, or a timing A2 after a little time has elapsed from the start of the attack. And an algorithm for detecting the timing A3 when the attack ends.

  Therefore, when the orbit correction of the uniform modulation reproduction position for attack reproduction as in the above embodiment is performed, the timing is shifted from an appropriate timing depending on the type of the musical instrument of the audio signal to be pitch converted and the attack detection algorithm. There is a possibility that the musical instrument sound after pitch conversion is reproduced at the timing.

  Therefore, in the present embodiment, the signal processing device is configured so that a desired direction and a desired offset amount can be designated according to, for example, operation of an operator provided in the karaoke device. Further, the trajectory correcting unit shifts the target reproduction timing Tx from the timing at which the reproduction reference position BA becomes the storage position of the attack part in the ring buffer 1 by an offset amount designated in the direction designated by the operation of the operator. The offset adjusting means is provided. In the case of use in a karaoke apparatus, the offset amount may be automatically set using music data (information such as genre and musical instrument data used). Alternatively, the music data may have offset amount information, and the offset amount may be set according to the offset amount information in the music data. In this case, time-series offset amount information may be included in the music data, and the setting of the offset amount may be repeated according to the offset amount information that is reproduced together with the music data and changes every moment.

According to the present embodiment, it is possible to optimize the playback timing of the attack portion of the audio signal after pitch conversion in accordance with the type of musical instrument and the attack detection algorithm.
When the present invention is applied to an apparatus that performs pitch conversion of a plurality of types of audio signals indicating performance sounds of different types of instruments for each channel, the modulation for attack reproduction after such trajectory correction is performed for each channel. The offset amount and the shift direction applied to the target position of the reproduction reference position that crosses the reproduction position may be designated by operating the operator.

<Third Embodiment>
Next, a third embodiment of the present invention will be described with reference to FIGS. In the first embodiment, as the pitch conversion ratio r increases, the time gradient of the modulation reproduction positions MA1 and MA2 increases. In this case, in order to prevent trouble in the playback of the attack part, the audio signal fades in corresponding to the attack playback modulation playback position and the audio signal fades out corresponding to the non-attack playback modulation playback position. It is necessary to shorten the time T1 (see FIG. 3). However, if the time T1 is shortened and fade-in and fade-out are performed, the sound quality of the output audio signal deteriorates. The present embodiment solves this problem.

  In the present embodiment, a sufficiently large address difference exceeding an address difference corresponding to one period of the modulated reproduction position MA1 (MA2) is provided between the read address of the attack detection unit 4 and the reproduction reference position BA.

  In the example shown in FIG. 7, the attack portion written in the ring buffer 1 is detected within the period T10 in which the modulation reproduction position MA1 is larger than the reproduction reference position BA. The detected attack portion is to be reproduced at timing Tx after the jump of the second modulation / reproduction position MA1 after the attack detection. Therefore, this timing Tx becomes the target reproduction timing. Therefore, the trajectory correcting unit 323 sets the modulation reproduction position MA1 jumping immediately before the target reproduction timing Tx as the attack reproduction modulation reproduction position and the other modulation reproduction position MA2 as the non-attack reproduction modulation reproduction position. Here, in the example shown in FIG. 7, the time between the target reproduction timing Tx of the attack portion and the jump timing of the modulation reproduction position MA1 immediately before the time exceeds a predetermined threshold value.

  Accordingly, the trajectory correcting unit 323 fades out the audio signal corresponding to the attack reproduction modulation reproduction position MA1 and fades the audio signal corresponding to the non-attack reproduction modulation reproduction position MA2 within the period T20 one cycle after the period T10. In is started (the start of time T0 in FIG. 3 is generated), and the same control as in the first embodiment is performed.

  In this case, since the time between the target reproduction timing Tx and the jump timing of the modulation reproduction position MA1 immediately before it is sufficiently long, the audio signal fades out and non-attack reproduction corresponding to the modulation reproduction position for attack reproduction within the period T20. Even if the audio signal fade-in corresponding to the modulation playback position for audio is started (time T0), then the audio signal corresponding to the modulation playback position for attack playback and the audio corresponding to the modulation playback position for non-attack playback are used. The time T1 for fading out the signal can be made sufficiently long.

  On the other hand, in the example shown in FIG. 8, the time between the target reproduction timing Tx of the attack portion and the jump timing of the modulation reproduction position MA1 immediately before it is shorter than a predetermined threshold value. In this case, the fade-out of the audio signal corresponding to the modulation reproduction position for attack reproduction and the fade-in of the audio signal corresponding to the modulation reproduction position for non-attack reproduction are performed within a period T20 one cycle after the period T10 in which the attack detection is performed. When started, it is difficult to make the time T1 for the fade-in of the audio signal corresponding to the subsequent modulation playback position for attack playback and the fade-out of the audio signal corresponding to the non-attack playback modulation playback position sufficiently long. It is.

  Therefore, the trajectory correcting unit 323 corrects the trajectories of the attack reproduction modulation reproduction position MA1 and the non-attack reproduction modulation reproduction position MA2 and controls the coefficients C1 and C2 as follows. First, the trajectory correcting unit 323 jumps the attack reproduction modulation reproduction position MA1 after the attack part is detected, and further jumps the non-attack reproduction modulation reproduction position MA2 after a period T15 shown in FIG. 8 elapses. Wait for.

  At the time when the modulation reproduction position MA2 for non-attack reproduction jumps, the coefficient C1 multiplied by the audio signal corresponding to the modulation reproduction position MA1 for attack reproduction is “1”, and the audio signal corresponding to the modulation reproduction position MA2 for non-attack reproduction. The coefficient C2 to be multiplied by “0” is “0”. Therefore, the trajectory correcting unit 323 takes a predetermined time T0, and signals the fade-out of the audio signal corresponding to the modulation reproduction position MA1 for attack reproduction and the fade-in of the audio signal corresponding to the modulation reproduction position MA2 for non-attack reproduction. The combining unit 35A is caused to execute.

  When the fade-out and the fade-in are completed, the trajectory correcting unit 323 jumps the attack reproduction modulation reproduction position MA1. At that time, the trajectory correcting unit 323 jumps the attack reproduction modulation reproduction position MA1 so that the trajectory of the attack reproduction modulation reproduction position MA1 after the trajectory correction crosses the reference reproduction position BA at the target reproduction timing Tx. Determine the position.

Then, the trajectory correction unit 323 takes a predetermined time T1 after the jump of the attack reproduction modulation reproduction position MA1, and fades in the audio signal corresponding to the attack reproduction modulation reproduction position MA1, and the non-attack reproduction modulation reproduction. The signal synthesis unit 35A is caused to execute fade-out of the audio signal corresponding to the position MA2.
Subsequent operations are the same as those in the first embodiment.

  In the example shown in FIG. 8, since the beginning of time T0 is advanced to the jump timing of the non-attack playback modulation playback position MA2, then the audio signal fade-in corresponding to the attack playback modulation playback position MA1 and the non-attack The length of the time T1 for executing the fade-out of the audio signal corresponding to the reproduction modulation reproduction position MA2 can be made sufficiently long.

  As described above, according to the present embodiment, even when the pitch conversion ratio r increases and the time gradient of the modulation reproduction positions MA1 and MA2 increases, the pitch of the output audio signal is not degraded. It is possible to reduce a deviation in attack timing accompanying conversion.

<Fourth embodiment>
FIG. 9 is a block diagram showing the configuration of the signal processing apparatus according to the fourth embodiment of the present invention. In this signal processing device, the reproduction unit 3A of the signal processing device (FIG. 1) according to the first embodiment is replaced with a reproduction unit 3B, and a basic period detection unit 5 is further added. In the reproduction unit 3B, the read control unit 32A in the first embodiment is replaced with a read control unit 32B.

  In this signal processing device, the fundamental period detector 5 detects the period of the fundamental component of the audio signal written to the ring buffer 1. Similar to the read control unit 32A of the first embodiment, the read control unit 32B includes a modulation reproduction position generation unit 321, a read address generation unit 322, and a trajectory correction unit 323 (see FIG. 2). However, as shown in FIG. 10, the modulation reproduction position generation unit 321 of the read control unit 32B has a period n times (n is an integer) the period τ of the fundamental component of the audio signal detected by the fundamental period detection unit 5. The modulation reproduction positions MA1 and MA2 are changed. As shown in FIG. 11, the trajectory correction unit 323 of the read control unit 32B causes the audio signal to fade out corresponding to the attack reproduction modulation reproduction position (MA1 in the example of FIG. 11) in response to the attack detection. When jumping the modulation reproduction position MA1 for attack reproduction from the current position Mst to the target position Men, the target position Men has a length α corresponding to the period of the fundamental wave component detected by the fundamental period detector 5 (that is, the fundamental wave component). The width of the address where the sample sequence for one wavelength is stored). Specifically, several possible lengths (m−1) α, mα, (m + 1) α,... That are integer multiples of the length α are assumed as candidates for the distance from the current position Mst to the target position Men. The target position of the jump destination is selected by selecting the one from which the orbit of the modulation reproduction position MA1 for attack reproduction after the jump crosses the orbit of the reproduction reference position BA is closest to the target reproduction timing Tx of the attack part. Men are determined.

  According to this aspect, although the playback timing of the attack part slightly deviates from the ideal playback timing, it is possible to align the phases of the two audio signals that are crossfade while the section in the vicinity of the attack part is played back. So you can play more natural music.

<Other embodiments>
The first to fourth embodiments of the present invention have been described above, but other embodiments are conceivable for the present invention. For example:

(1) In each of the above embodiments, the coefficients C1 and C2 that change sinusoidally are used for the crossfade process, and the square sum of the coefficients C1 and C2 is always 1. When such coefficients C1 and C2 are used for the cross-fade process, there is an effect that both audio signals can be smoothly connected even when the waveform of the two audio signals to be processed has no periodicity. can get. However, when such coefficients C1 and C2 are used for the cross-fade process, there arises a problem that the level of the output audio signal varies when the waveform of the audio signal to be processed has periodicity. Therefore, the signal processing device may be configured so that the waveforms of the coefficients C1 and C2 can be switched to another waveform. For example, when a means for analyzing the periodicity of the audio signal to be processed is provided and a determination result indicating that there is periodicity is obtained, the coefficient C1 for changing the coefficients C1 and C2 for the crossfade processing into a triangular wave shape And C2 and the sum of the coefficients C1 and C2 may always be 1. When such coefficients C1 and C2 are used for the cross-fade processing, both audio signals can be smoothly connected when the waveforms of two audio signals to be processed have periodicity.

(2) In each of the above embodiments, a signal processing apparatus that performs pitch conversion is described. However, the present invention is naturally applicable to a signal processing apparatus that performs only tempo conversion without performing pitch conversion.

DESCRIPTION OF SYMBOLS 1 ... Ring buffer, 2 ... Write part, 3A, 3B ... Reproduction | regeneration part, 4 ... Attack detection part, 31 ... Reproduction | regeneration reference position update part, 32A, 32B ... Read control part, 33, 34 ... Interpolation part, 35A ... Signal Combining unit, 321... Modulation reproduction position generating unit, 322... Read address generating unit, 323... Trajectory correcting unit, 354... Coefficient generating unit, 351, 352.

Claims (3)

A buffer for storing audio signals;
Writing means for writing an audio signal to the buffer while sequentially changing the write address of the buffer;
Attack detection means for detecting an attack portion appearing in the audio signal written to the buffer;
Reproduction means for generating an output audio signal by reading the audio signal from the buffer;
The reproducing means includes
Reproduction reference position update means for updating the reproduction reference position so as to follow the change of the write address of the buffer with a delay by a predetermined value;
While periodically repeating a change in a second time gradient different from the first time gradient that is the time gradient of the reproduction reference position and a jump in the direction opposite to the change direction in the second time gradient, A first modulation reproduction position that changes with the first time gradient as a whole is generated, and a second modulation reproduction position that is out of phase with respect to the first modulation reproduction position is generated. An output audio signal for reproducing two audio signals corresponding to the first and second modulation reproduction positions based on the audio signal stored in the buffer and generating the output audio signal using the two audio signals Generating means,
The output audio signal generation means includes
A means for generating the output audio signal by subjecting the two audio signals to cross fading processing, and corresponding to the first modulation / reproduction position before the first modulation / reproduction position jumps. The audio signal fades out and the audio signal corresponding to the second modulation / reproduction position fades in, and the audio signal corresponding to the second modulation / reproduction position jumps before the second modulation / reproduction position jumps. Signal synthesizing means for performing fade-in and fading-in of the audio signal corresponding to the first modulated reproduction position;
In response to the detection of the attack part by the attack detection means, the timing near the timing at which the reproduction reference position updated by the reproduction reference position update means becomes the storage position of the attack part of the audio signal in the buffer. Of the first modulation reproduction position and the second modulation reproduction position as an attack reproduction modulation reproduction position, the other as a non-attack reproduction modulation reproduction position, and the attack after the trajectory correction. The attack reproduction modulation reproduction position trajectory correction is performed so that the reproduction modulation reproduction position trajectory crosses the reproduction reference position trajectory at the target reproduction timing, and a predetermined period sandwiching the target reproduction timing in the middle over the, without executing the cross-fade process on said signal combining means, again the attack The audio signal corresponding to the use modulated playback position; and a trajectory correction means for outputting as the output audio signal,
The trajectory correcting means is
When the time between the target playback timing and the jump timing of the immediately preceding modulated playback position for attack playback is longer than a predetermined threshold, the attack playback is performed using the predetermined time before the jump timing. Causing the signal synthesizing unit to execute fade-out of the audio signal corresponding to the modulation reproduction position for audio and fade-in of the audio signal corresponding to the modulation reproduction position for non-attack reproduction,
When the time between the target reproduction timing and the jump timing of the attack reproduction modulation reproduction position immediately before is shorter than a predetermined threshold, the jump of the non-attack reproduction modulation reproduction position before the jump timing Using the predetermined time after the timing, the signal synthesizing unit executes the fade-out of the audio signal corresponding to the modulated playback position for attack playback and the fade-in of the audio signal corresponding to the modulated playback position for non-attack playback. ,
After the completion of the fade-out and fade-in, the attack reproduction modulation reproduction position is caused to jump, and at this time, the attack reproduction modulation reproduction position trajectory crosses the reproduction reference position trajectory at the target reproduction timing. , Determine the jump destination value of the attack reproduction modulation reproduction position,
After the jump of the attack playback modulation playback position, the fade-in of the audio signal corresponding to the attack playback modulation playback position and the non-attack playback modulation so that the fade-in and fade-out are completed before the target playback timing. Let the signal synthesis means execute the fade-out of the audio signal corresponding to the playback position,
When a predetermined time has elapsed from the target playback timing, the non-attack playback modulation playback position is jumped, and at that time, the non-attack playback modulation playback position jumps at the timing at which the attack playback modulation playback position jumps. Determine the jump destination value of the non-attack playback modulation playback position so that the trajectory of the attack playback modulation playback position crosses the trajectory of the playback reference position,
A fade-out of an audio signal corresponding to the modulated playback position for attack playback is performed using the period from the jump of the modulated playback position for non-attack playback to the jump of the modulated playback position for attack playback after trajectory correction. And a signal processing unit that causes the signal synthesizing unit to execute fade-in of an audio signal corresponding to the modulation reproduction position for non-attack reproduction. Comprising fundamental period detecting means for detecting the period of the fundamental component of the audio signal written to the buffer;
The trajectory correcting means jumps the attack reproduction modulation reproduction position from the current position to the target position so that the trajectory of the attack reproduction modulation reproduction position crosses the reproduction reference position trajectory at the target reproduction timing. 2. The target position and the target reproduction timing are determined such that a length from the current position to the target position is an integral multiple of a length corresponding to a period of a fundamental wave component. A signal processing device according to 1. An offset adjusting means for adjusting an offset of the target reproduction timing with respect to a timing at which the reproduction reference position updated by the reproduction reference position updating means becomes a storage position of an attack portion of an audio signal in the buffer. Item 3. The signal processing device according to Item 1 or 2.

Images