JP2005195968A - Pitch converting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a pitch converting device capable of converting the pitch of an audio signal such as a speech signal and a musical sound signal while leaving natural fluctuations with simple constitution in such a case. <P>SOLUTION: A desired section is cut out of sampling data stored in a ring memory as a phoneme based upon cycles corresponding to a pitch detected by a pitch detecting means 30, and the cut-out section is read out with time at a readout speed corresponding to a formant change coefficient FORMANT-VR; and the pitch detected by the pitch detecting means 30 is smoothed by a pitch smoothing means 32 and the phoneme is put together with the smoothed pitch in cycles corresponding to target pitch information SPITCH. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a pitch converter for converting the pitch of an audio signal, a musical sound signal, or the like.

  Conventionally, various pitch converters have been proposed. For example, in Japanese Patent No. 3329635 (Patent Document 1), an input audio signal is stored in a memory, a pitch (period) of the audio signal is detected, and a section corresponding to the detected period is stored in the memory. An effect device is disclosed in which a pitch is converted and reproduced by synthesizing the segmented segments in a cycle according to the newly designated pitch. In the present specification, a newly designated pitch is referred to as a target pitch, and this target pitch refers to an absolute pitch that is not related to the pitch of the input audio signal. As a method of designating the pitch, there is a method of designating by a relative value with respect to the pitch of the input audio signal, for the purpose of distinguishing from this method.

In addition, in Japanese Patent No. 3287230 (Patent Document 2), information specifying the pitch of a vocal sound to be generated by a singer and information specifying the pitch of a harmony sound to be given to the vocal sound are input, and singing is performed. Detects the pitch of the vocal sound uttered by the person, finds the ratio of the pitch to the pitch of the vocal sound that the singer should generate, and gives the vocal sound the information obtained by compressing the ratio A chorus effect device is disclosed in which the pitch of a harmony sound is converted by adding to the information specifying the pitch of the harmony sound. Therefore, in this apparatus, when the pitch of the vocal sound uttered by the singer is fluctuating, the fluctuation of the singer's pitch is extracted by the ratio of the pitch of the singer and the pitch of the vocal sound to be uttered. The added fluctuation is added to the harmony sound.
Japanese Patent No. 3329635 (FIG. 10 etc.) Japanese Patent No. 3287230

  However, the effect device described in Patent Document 1 is configured to synthesize the clipped audio signal at a period corresponding to the newly designated target pitch, and thus the converted sound is constant. It was generated at the pitch and sounded as if the robot was speaking. In particular, in the case of a chorus adding device used as an effect device in karaoke or the like, the pitch of the added chorus is constant, and thus there is a problem that it is unnatural and uncomfortable.

  Moreover, in the chorus effect imparting device described in Patent Document 2, the ratio of the pitch of the vocal sound that the singer should generate and the pitch of the vocal sound that the singer generated is obtained, and the harmony sound is calculated based on this ratio. This gives a problem that information specifying the pitch of the vocal sound that the singer should generate is necessary.

  The present invention has been made to solve the above-described problems. When converting the pitch of an audio signal such as an audio signal or a musical tone signal, the pitch of the audio is converted with a simple configuration while leaving natural fluctuations. An object of the present invention is to provide a pitch conversion device that can perform the above-described operation.

In order to achieve the above object, a pitch conversion apparatus according to claim 1 of the present invention includes an input means for inputting an audio signal, and a pitch detection means for sequentially detecting the pitch of the audio signal input to the input means. The pitch detection means for smoothing the pitch detected by the pitch detection means, the pitch smoothing means for obtaining smoothed pitch information, the target pitch specifying means for specifying the target pitch, and the pitch of the audio signal input to the input means are converted. Pitch conversion means; and control means for controlling the pitch conversion means based on the smoothed pitch information smoothed by the pitch smoothing means and the target pitch specified by the target pitch specifying means.

  According to the pitch conversion apparatus of the first aspect, the pitch converting apparatus includes a smoothing unit that sequentially detects the pitch of the input audio signal and smoothes the pitch, and the smoothed pitch and the designated target pitch are set. Based on this, pitch conversion is performed.

  The pitch converter according to claim 2 is the pitch converter according to claim 1, wherein the smoothing means performs low-pass filter processing on the sequentially detected pitches.

  The pitch conversion device according to claim 3 is the pitch conversion device according to claim 1 or 2, wherein the smoothing means subtracts a previously smoothed value from a pitch value detected this time by the pitch detection means. Smoothing pitch information is obtained by multiplying the difference value by a predetermined coefficient.

  The pitch conversion device according to claim 4 is the pitch conversion device according to any one of claims 1 to 3, wherein the pitch conversion unit is configured to perform predetermined processing of the audio signal based on the pitch detected by the pitch detection unit. The section is cut out, and the cut out section is synthesized with a period according to the smoothed pitch information smoothed by the pitch smoothing means and the target pitch specified by the target pitch specifying means.

  The pitch conversion device according to claim 5 is the pitch conversion device according to any one of claims 1 to 3, wherein the front pitch conversion means smoothes the audio signal stored in the waveform storage means by the pitch smoothing means. Data is read at a speed corresponding to the smoothed pitch information and the target pitch specified by the target pitch specifying means.

  According to the pitch conversion device of the first aspect, pitch smoothing means for sequentially detecting the pitch of the input audio signal and smoothing the pitch is provided, and based on the smoothed pitch and the designated target pitch. Therefore, the pitch of the input audio signal is converted to the specified target pitch, and if the pitch of the input audio signal fluctuates, the pitch of the converted audio signal also fluctuates. There is an effect that can be made.

  According to the pitch conversion device of claim 2, in addition to the effect of the pitch conversion device of claim 1, the pitch smoothing means performs low-pass filter processing on the sequentially detected pitch, There is an effect that it can be smoothed by a simple process.

  According to a third aspect of the present invention, in addition to the effect produced by the first aspect of the present invention, the pitch smoothing means is previously smoothed from the pitch value detected this time by the pitch detecting means. Since the smoothed pitch information is obtained by multiplying the difference value obtained by subtracting the obtained value by a predetermined coefficient, there is an effect that the smoothing process can be performed by a simple calculation.

  According to a fourth aspect of the present invention, in addition to the effect produced by the pitch conversion device according to any one of the first to third aspects, the pitch conversion unit is configured to generate the pitch conversion unit based on the pitch detected by the pitch detection unit. A predetermined section of the audio signal is cut out, and the cut out section is synthesized with a period according to the smoothed pitch information smoothed by the pitch smoothing means and the target pitch specified by the target pitch specifying means. There is an effect that only the pitch can be converted without changing the formant of the input audio signal.

  The pitch conversion device according to claim 5 is the pitch conversion device according to any one of claims 1 to 3, wherein the previous pitch conversion means smoothes the audio signal stored in the waveform storage means by the pitch smoothing means. The smoothed pitch information and the target pitch designated by the target pitch designation means are read out at a speed, so that the formant of the input audio signal is converted at the same time, but the configuration is simple. There is.

  As shown in FIG. 1 (a), the first embodiment of the present invention has an input terminal 2 to which an audio signal (audible frequency signal) such as an analog musical tone signal or a voice signal is input. The audio signal supplied to 2 is converted into a digital audio signal (sampling data) by the A / D converter 4. Between the A / D converter 4 and the input terminal 2, the audio signal is limited to a frequency that is 1/2 or less of the sampling frequency (for example, 48 kHz) in the A / D converter 4 to prevent aliasing. For this purpose, a low-pass filter 6 is provided.

  The sampling data digitized by the A / D converter 4 is supplied to a DSP (digital signal processing device) 8 and is supplied from the DSP 8 to the RAM 12. The RAM 12 is used as a ring memory that sequentially stores input sampling data.

  Next, the DSP 8 reads the sampling data temporarily stored in the ring memory, performs processing, and supplies it to the D / A converter 14. The D / A converter 14 converts the sampling data processed by the DSP 8 into an analog audio signal and supplies the analog audio signal to the low-pass filter 16. This analog audio signal is output from the output terminal 18 after unnecessary signal components are removed by the low-pass filter 16, amplified by an amplifier (not shown), and emitted from a speaker or the like.

  The DSP 8 processes sampling data according to a preset program. For this process, parameters such as formant change coefficient FORMAT-VR and target pitch information SPITCH set by the operator 28 operated by the user are used. These formant change coefficient FORMANT-VR and target pitch information SPITCH are provided with volume knobs for setting each of them, and the values set by these volume knobs are detected by the CPU 22, supplied to the DSP 8, and stored in the DSP 8. Stored in a register. The target pitch information SPITCH is a period corresponding to the target pitch (frequency), and is the reciprocal of the target pitch.

  A ROM 24 and a RAM 26 storing programs and the like are connected to the CPU 22 via a bus, and the pitch conversion device is controlled by executing the programs stored in the ROM 24. As the processing executed by the CPU 22, as described above, the value set by the operation element 28 is detected, and the DSP 8 is controlled such that the parameter corresponding to the setting is supplied to the DSP 8, and the set parameter is displayed on the display ( There is a process such as displaying (not shown).

  FIG. 1B is a functional block diagram showing the functions executed by the DSP 8, and shows a basic embodiment of the present invention. In this embodiment, there is a functional block for sequentially storing the sampling data input to the input terminal 23 in the ring memory of the RAM 12, but for the sake of easy explanation, the functional block is shown in FIG. Is omitted.

  In the basic embodiment shown in FIG. 2B, an input terminal 23, a pitch conversion means 36, a pitch detection means 30, a pitch smoothing means 32, a control means 34, and an output terminal 29 are provided. .

  The pitch detection means 30 detects the zero cross points from the sampling data, and sequentially detects periods in which the intervals between the zero cross points are substantially the same. The pitch smoothing unit 32 performs smoothing by sequentially inputting the periods detected by the pitch detecting unit 30. When the pitch of the input audio signal is fluctuating, the fluctuation is removed, and a smoothed (averaged) pitch (period) is obtained.

  The control means 34 includes a pitch detection signal from the pitch detection means 30 and smoothed pitch information lp smoothed by the pitch smoothing means 32, and a formant change coefficient FORMAT-VR set by the operation of the operator 28 via the CPU 22. The pitch conversion means 36 is controlled using parameters such as target pitch information SPITCH.

  The pitch converting means 36 cuts out a desired section as a phoneme based on the period corresponding to the pitch detected by the pitch detecting means 30 from the sampling data stored in the ring memory, and cuts out based on the control of the control means 34. The read section is read out at a speed corresponding to a desired formant and synthesized at a period corresponding to the reproduction pitch information WIDTH.

The pitch smoothing means 32 smoothes (averages) the period corresponding to the pitch of the audio signal sequentially detected by the pitch detecting means 30 as described above, and a low-pass filter (low-pass filter) can be used. . As an example of the low-pass filter, if the detected pitch (period) is PITCH and the smoothing rate is α (a coefficient having a predetermined value larger than 0 and smaller than 1.0), the smoothed smoothing pitch lp (smoothing pitch) Information)
lp = (PITCH−lp) × α + lp (1)
Can be obtained as

  Here, if the value of the smoothing rate α is 1.0, lp becomes the value of the detected pitch that is not smoothed, and if the value of the smoothing rate α is set to a value close to 0, the value of the detected pitch The smoothed pitch information does not follow the change. By setting the smoothing rate to an appropriate value, when the pitch of the input audio signal fluctuates, the pitch from which the fluctuation is removed, in other words, the center pitch is obtained.

  The control means 34 controls the pitch converting means 36 so as to convert the pitch from which the fluctuation is removed into a designated target pitch. As a result, the pitch fluctuation of the input audio signal is retained in the pitch-converted audio signal.

  In this basic embodiment, the phoneme is reproduced by using a processing path for reproducing the first waveform and a processing path for reproducing the second waveform. The phonemes are reproduced with a period twice as long as the intended period, and these are synthesized.

  The effect adding operation will be described with reference to the flowcharts shown in FIGS. The DSP 8 includes a register PITCH that stores a period corresponding to the detected pitch of the sampling data, and a register SADRS that stores an address for reading (reading) the sampling data from the ring memory. Further, a register PHASE for counting whether a reproduction pitch period length described later has been reached, a register PH1 for counting the phase of the first waveform, and a register PH2 for counting the phase of the second waveform are also provided. ing.

  A register WIDTH for storing reproduction pitch information obtained by multiplying the period PITCH detected by the pitch detection means 30 by the target pitch information SPITCH and dividing the product by the smoothed pitch information lp, the detected period And a register LENGTH for storing the length of the envelope determined from the formant coefficient FORMANT-VR, a register lp for storing smoothed pitch information which is a smoothed pitch, and a register ENV1 for storing the envelope of the first waveform And a register ENV2 for storing the envelope of the second waveform.

  Further, the step rate of WINDOW1 and WINDOW2 determined based on the values of the register WINDOW1 for determining the shape of the envelope of the first waveform, the registers WINDOW2 and LENGTH for determining the shape of the envelope of the second waveform, Register W-RATE to be stored, register SADRS1 to store the first waveform cut-out start address, register SADRS2 to store the second waveform cut-out start address, determination of the first and second waveform cut-out start positions, etc. A flag F used for the purpose is also provided.

  These are initialized when the power is turned on. That is, the flag F is set to 1 and the others are set to 0. In the following description, it is assumed that appropriate values are already stored in each register. Each step of the flowcharts shown in FIGS. 2 and 3 is executed each time sampling data is input from the A / D converter 4 to the DSP 8.

  In FIG. 2, when sampling data is supplied from the A / D converter 4, it is written to the ring memory (S2). Next, pitch detection processing is performed based on the input sampling data (S4). The process of S4 corresponds to the pitch detection means 30. This pitch detection process outputs SADRS, which is the address of the zero cross position, and PITCH, which is a cycle corresponding to the detected pitch, when the pitch is detected.

  This pitch detection process sequentially compares time intervals between adjacent zero crosses, for example. For example, FIG. 5A shows a waveform of an input audio signal, and the horizontal axis represents time and the vertical axis represents the amplitude value of the waveform. In this figure, it is assumed that the time intervals between the zero crosses of the input signal are a0, b0, c0, d0, a1, b1, c1, and d1. The time interval a0 between the first zero crosses is compared with the time interval b0 between the next zero crosses. If they are different, the time interval (a0 + b0) obtained by adding them is compared with the adjacent time interval (c0 + d0). . If both are different in this comparison, the time interval (a0 + b0 + c0) is compared with the adjacent time interval (d0 + a1 + b1). If they are different, the time interval (a0 + b0 + c0 + d0) is compared with the adjacent time interval (a1 + b1 + c1 + d1). When the two values are almost the same in this comparison, the time interval (a1 + b1 + c1 + d1) is output as PITCH, and the address corresponding to time t0 is output as the address at the zero cross position.

  In addition, as disclosed in, for example, Japanese Patent Laid-Open No. 3-288200, the zero cross position of the sampling data and the peak position of the waveform signal are detected, and the time interval between the zero cross position and the peak position is previously detected. It is also possible to use one that detects the pitch by comparing the time interval with the zero cross position.

  It is determined whether or not pitch detection has been performed in the process of S4 (S6). If the pitch is detected (S6: Yes), the zero-cross position detected in the process of S4 is set as the cut-out address, and thereby the register SADRS The value is updated, and the detected pitch is set to one cycle length, thereby updating the value of PITCH (S8).

  When the pitch detection is not performed (S6: No), or when the pitch detection is performed and both the above registers are updated, the values of the registers PHASE, PH1, and PH2 are respectively advanced by 1 (S10). .

  Next, it is determined whether the register PHASE is equal to or greater than the value of WIDTH (S12). WIDTH stores one cycle length corresponding to the pitch to be reproduced as will be described later, and determines whether or not the value of PHASE has reached this WIDTH value. That is, it is determined whether or not the PHASE value has reached a predetermined reproduction pitch. The value of WIDTH corresponds to reproduction pitch information.

  If the PHASE value does not reach the WIDTH value (S12: No), the process proceeds to the waveform processing in S32 of the flowchart shown in FIG. If the value of PHASE has reached the value of WIDTH (S12: Yes), the value of register PHASE is set to 0 (S14).

  Next, a pitch smoothing process is performed. In this process, a value obtained by subtracting the value of the register lp from the value of the register PITCH is multiplied by the coefficient α, and a value obtained by adding the value of the register lp to the value is used as a new lp. The values sequentially stored in are smoothed. The smoothed value lp is stored in the register lp (S16).

  In order to determine a new WIDTH value, the PITCH value is multiplied by the target pitch information SPITCH, and the product is divided by the smoothed pitch information lp value. This is stored in the register WIDTH, and the PITCH value is divided by the formant coefficient FORMANT-VR and stored in the register LENGTH to determine the envelope periods of the first waveform and the second waveform (S18). ).

  Next, proceeding to the flowchart shown in FIG. 3, it is determined whether or not the value of the register LENGTH is larger than the value of the register WIDTH (S20). If the LENGTH value is larger than the WIDTH value (S20: Yes), the LENGTH value is set as the WIDTH value (S22). This is to prevent the LENGTH value from exceeding the WIDTH value. In other words, since only two processing paths for adding waveforms are prepared in the waveform readout process of S32 described later, two or more waveforms are prevented from overlapping.

  If the LENGTH value is less than or equal to the WIDTH value, the process of S24 is performed without performing the process of S22. Further, following the process of S22, the process of S24 is also performed. In the process of S24, the reciprocal of the value of LENGTH is obtained and stored in the register W-RATE. This W-RATE value is used to step up the values of WINDOW1 and WINDOW2 as will be described later. Further, in the process of S24, the value of the flag F is also inverted from the current value.

  Next, it is determined whether the value of the flag F inverted in the process of S24 is 1 or -1 (S26). When the flag F is 1 (S26: Yes), in order to start reading the first waveform, the registers PH1 and WINDOW1 are set to 0, and the value of SADRS (cutout address) determined in the process of S8 is set to the cutout start address register. The data is stored in SADRS1 (S28).

  If the value of F inverted in the process of S24 is -1 (S26: No), the value of the register SADRS determined in S8 is set to 0 by setting the registers PH2 and WINDOW2 to 0 in order to start reading the second waveform. The (cutout address) is stored in the cutout start address register SADRS2 (S30).

  By such steps S12, S14, S16, S18, S20, S22, S24, S26, S28, and S30, the first and second time intervals corresponding to the period length of two playback pitches (twice the value of WIDTH) are passed. The cut start address of the second waveform data is updated, and the value of the flag F is inverted every time the PHASE value reaches the WIDTH value.

  Following the processing of S28 or S30, or when it is determined in the processing of S12 that the value of the register PHASE has not reached the value of the register WIDTH, a waveform reading process is performed (S32). In this manner, the waveform reading process is performed without changing the WIDTH value and LENGTH value (and thus the W-RATE value) until the PHASE value reaches WIDTH.

  In the waveform readout process described later, the values of envelopes ENV1 and ENV2 of the first and second waveforms are controlled by the value of W-RATE.

  Since the value of W-RATE, and hence LENGTH, is obtained by changing the detected pitch by the formant change coefficient FORMAT-VR in S18, the pitch of the waveform to be subjected to the waveform reading process corresponds to the formant change coefficient FORMANT-VR. It will be a thing.

  Next, the waveform reading process will be described with reference to FIG. First, the value of the counter WINDOW1 is incremented by the value of W-RATE (S34). Then, it is determined whether the value of the incremented WINDOW1 is less than 1, 1 or more, less than 2, or 2 or more (S36).

  If it is smaller than 1, the value of WINDOW1 is stored in the register ENV1 (S38). If it is 1 or more and smaller than 2, the value obtained by subtracting the value of WINDOW1 from 2 is stored in the register ENV1 (S40). At this time, the value of ENV1 is set to 0 (S42).

  In the processing of S34 to S40, a sawtooth wave whose value increases by the value of W-RATE is created, and this value is folded back to 1, thereby creating ENV1. However, when the value of WINDOW1 exceeds 2, ENV1 is set to 0 by the process of S42. That is, a triangular wave that increases to 1 by W-RATE, which is the reciprocal of the value of LENGTH determined based on the formant coefficient FORMANT-VR and the detection pitch, and then decreases to 0 by W-RATE is converted into an envelope of the first waveform. It is created as.

  Further, following the processing of S38, S40 or S42, a value obtained by multiplying the value of the register PH1 (stepping value of the cut-out address) by the formant coefficient FORMAT-VR is stored as the cut-out start address of the first waveform. The value is added to the value of the register SADRS1, and is stored in the register ADRS1 that stores the cut-out address of the first waveform (S44). Subsequently, the waveform data DATA1 of the first waveform is read from the ring memory with the cut-out address ADRS1 (S46).

  As described above, since the read address is changed by the formant coefficient FORMAT-VR, as a result, the read speed of the waveform data DATA1 is changed by the formant coefficient FORMAT-VR.

  Following this, the value of WINDOW2 is incremented by W-RATE (S48). It is determined whether the value of the incremented WINDOW2 is less than 1, 1 or more and less than 2, or 2 or more (S50). If the value of WINDOW2 is smaller than 1, the value of WINDOW2 is stored in the register ENV2 (S52). If the value of WINDOW2 is 1 or more and less than 2, the value obtained by subtracting the value of WINDOW2 from 2 is stored in the register ENV2. If the value of WINDOW2 is 2 or more, the value of ENV2 is set to 0 (S56). In this way, the envelope of the second waveform is prepared.

  Subsequent to the processing of S52, S54, or S56, the value obtained by multiplying the value of the register PH2 by the value of the formant coefficient FORMAT-VR and the value of the extraction start address SADRS2 for the second waveform data is added to the second waveform. It is stored in the register ADRS2 for the cut-out address for data (S58). Then, the waveform data (DATA2) of the second waveform is read from the ring memory at the address ADRS2 (S60).

  An output OUT is obtained by adding the value of ENV1 multiplied by the value of DATA1 read out in this manner and the value of DATA2 multiplied by the value of ENV2 (S62). After such waveform reading processing is performed, output OUT is sent out as shown in FIG. 3 (S64).

  FIG. 5 shows the waveform of each part when WIDTH set according to the target pitch information SPITCH is set longer than the detected period PITCH and the formant change coefficient FORMAT-VR is set to 1. That is, FIG. 5 shows a case where the pitch of the input audio signal is converted to a low value and the formant characteristics remain unchanged.

FIG. 5A shows an input waveform, which is one cycle from time t0 to t1, and that cycle is P0, and then from time t1 to t2 is the next one cycle, which is shown as P1. Yes.
The same applies hereinafter.
(B) is a change in the value of the register PHASE, and is a sawtooth wave having a period of WIDTH, which is a period corresponding to the pitch of the reproduced musical sound. The times for each cycle are tw0, tw1,.
(C) is a diagram showing PH1, and (d) is a diagram showing PH2, and is a sawtooth wave having a period twice that of WIDTH. PH1 and PH2 are different in phase by WIDTH.

  (E) is a diagram showing ENV1, and LENGTH is set to a cycle corresponding to the pitch of the input audio signal. One period of the audio signal input during this LENGTH period is read out at the same period as the input sampling period.

When the time is tw2, the period of P0 of the input waveform is determined. At this time, the value of LENGTH is set to P0, and the cut-out address SADRS is set to the address stored in the ring memory at time t0.
(F) is a diagram showing ENV2, and LENGTH is set to a cycle corresponding to the pitch of the input audio signal as in ENV1. ENV1 and ENV2 are envelope waveforms that are different in phase by WIDTH and multiplied by the read waveform, and the envelopes are cross-faded between ENV1 and ENV2.

  (G) and (h) show how the envelopes ENV1 and ENV2 are multiplied by the read phonemes, and these waveforms are synthesized and output.

  As described above, in this embodiment, the pitch of the output signal is multiplied by the pitch detected by the pitch detection means PITCH and the set target pitch information SPITCH, and the product is smoothed pitch information. It is determined by the value of WIDTH, which is a value divided by lp, and the read speed of the output signal is determined by the formant change coefficient FORMAT-VR.

  Therefore, when the pitch of the input audio signal fluctuates, the fluctuation of the pitch of the output signal corresponding to the smoothing rate is maintained while the center pitch is converted to the target pitch.

  FIG. 6 is a block diagram illustrating an example in which the present invention is applied to a chorus effect device for explaining a second embodiment of the present invention. The difference from FIG. 1 (b) is that the pitch converting means includes four chorus channels 36a, 36b, 36c, and 36d. The other pitch detection means 30 and pitch smoothing means 32 are the same, and the control means 34 supplies the detected pitch information PITCH and the respective reproduction pitch information WIDTH to each chorus channel.

  For example, when this chorus effect device is applied to a karaoke device, pitch information corresponding to a plurality of harmonies is supplied from a sequencer provided in the karaoke device according to the progress of a song by a note number defined by the MIDI standard. The The CPU 22 obtains target pitch information SPITCH corresponding to these note numbers and supplies it to the control means 34. The control means 34 obtains reproduction pitch information WIDTH corresponding to the detected pitch PITCH and the smoothed pitch lp for each target pitch information SPITCH, and supplies it to each chorus channel.

  In addition, the formant can be changed by changing and setting the formants of a plurality of harmonies. For example, if the input vocal is a male voice, if the harmony formant is set to female or child formant, the effect is as if a female or child harmony was added to the male vocal. Can do.

  FIG. 7 is a block diagram of an example in which the present invention is applied to an electronic musical instrument for explaining a third embodiment of the present invention. The difference from FIG. 1A is that an audio signal is stored in advance in a storage 38 which is a storage means, and a means for designating a target pitch is a keyboard 40. The other configurations are as shown in FIG. It is the same as (a) and has the same part number. FIG. 1A shows an embodiment in which an audio signal is input from the outside and the pitch of the audio signal is converted. In the case of FIG. 7, a large number of samples are sampled in a large-capacity memory such as a hard disk. The waveform is stored, one of the waveforms is selected, and the target pitch is designated by pressing a key provided on the keyboard 40.

  In this case, the waveform selected from the waveforms stored in the storage 38 is read out, the pitches are sequentially detected and the pitches are smoothed, and the smoothed smoothed pitches and the keyboard 40 are designated. Pitch conversion is performed based on the target pitch. As a result, fluctuations in the pitch of the waveform stored in the storage 38 can be added to the pitch of the waveform that is output after being pitch-converted.

  In the above embodiment, the pitch conversion device is a method of cutting out a predetermined section of the audio signal and synthesizing the cut out section with a period corresponding to the reproduction pitch. The pitch may be converted by sampling at a frequency and storing it, and reading at a sampling frequency different from the sampling frequency. For example, the chorus effect device disclosed in Japanese Patent No. 2519441 adopts this method. In this publication, when the target pitch is f1, the pitch of the input audio signal is f2, the pitch conversion amount P is f1 / f2, and the input audio signal is stored in the memory at a predetermined sampling frequency F1 and read out. The pitch conversion amount P is read at a sampling frequency F2 obtained by multiplying F1. In order to apply the present invention to this method, the detected pitch f2 may be replaced with a pitch smoothed by the pitch smoothing means 32. In this example, the so-called variable sampling method is used to change the sampling frequency. However, the sampling frequency is constant, the width of the address that advances is set to the width corresponding to the reproduction speed, and the sampling point is the decimal point of the address. In this case, a fixed sampling method may be used in which an amplitude value corresponding to the decimal point address is obtained by an interpolation method.

  The input means described in claim 1 corresponds to the input terminal 23 in FIG. 1B, and the target pitch designating means is the operation element 28 in FIG. 1A or the MIDI described with reference to FIG. This corresponds to the information and the keyboard 40 shown in FIG.

  The present invention has been described above based on the embodiments. However, the present invention is not limited to the above-described embodiments, and various modifications and changes can be easily made without departing from the spirit of the present invention. Can be inferred.

  For example, in the above embodiment, the pitch detection means 30 is provided in the DSP 8, but this processing may be executed by the CPU 22. In that case, the DSP 8 sends time information between zero crosses, the peak time of the waveform, and the like to the CPU 22, and the CPU 22 obtains autocorrelation from these values by calculation and extracts the period.

  Similarly, in the above embodiment, the pitch smoothing means 32 is provided in the DSP 8, but this processing may be executed by the CPU 22. In that case, the CPU 22 may read the value of the pitch detected by the pitch detection means provided in the DSP 8 and supply the smoothed smoothed pitch information to the control means 34. Similarly, the process performed by the control unit 34 may be executed by the CPU 22.

Moreover, although the arithmetic expression (1) of the low-pass filter which smoothes the value of the pitch detected sequentially is illustrated, as an arithmetic expression instead of the above,
lp = PITCH × α + (1−α) × lp
Or other smoothing calculation methods may be used.

(A) is a block diagram of the pitch conversion apparatus by this invention, (b) is a block diagram for demonstrating the process in DSP. It is a flowchart which shows a part of process which DSP performs. It is a flowchart following the flowchart shown in FIG. It is a flowchart which shows the detail of the waveform read-out process in the flowchart of FIG. It is a figure which shows the waveform for every part of the process performed in DSP. It is a figure which shows the block diagram of the chorus effect apparatus to which this invention is applied. It is a figure which shows the block diagram of the electronic musical instrument to which this invention is applied.

Explanation of symbols

8 DSP
22 CPU
23 Input terminal (input means)
30 Pitch detection means 32 Pitch smoothing means 34 Control means 36 Pitch conversion means

Claims (5)

An input means for inputting an audio signal;
Pitch detection means for sequentially detecting the pitch of the audio signal input to the input means;
Pitch smoothing means for smoothing the pitch detected by the pitch detecting means and obtaining smoothed pitch information;
Target pitch designating means for designating the target pitch;
The pitch conversion means for converting the pitch of the audio signal input to the input means, the pitch conversion based on the smoothed pitch information smoothed by the pitch smoothing means and the target pitch specified by the target pitch specifying means. And a control means for controlling the means.   2. The pitch converting apparatus according to claim 1, wherein the pitch smoothing means performs low-pass filter processing on the pitches sequentially detected by the pitch detecting means.   The pitch smoothing means obtains smoothed pitch information by multiplying a difference value obtained by subtracting a previously smoothed value from the pitch value detected this time by the pitch detecting means by a predetermined coefficient. The pitch converter according to claim 1 or 2, characterized in that   The pitch converting means cuts out a predetermined section of the audio signal based on the pitch detected by the pitch detecting means, and smoothed pitch information smoothed by the pitch smoothing means and the target pitch. 4. The pitch conversion apparatus according to claim 1, wherein the pitch conversion apparatus is one that synthesizes at a period corresponding to the target pitch designated by the designation means. The previous pitch conversion means reads the audio signal stored in the waveform storage means at a speed corresponding to the smoothed pitch information smoothed by the pitch smoothing means and the target pitch specified by the target pitch specifying means. The pitch converter according to any one of claims 1 to 3, wherein:

Images