JP3149466B2 - Pitch extraction device and electronic musical instrument using the same - Google Patents

Abstract

PURPOSE:To exactly detect the frequencies whatever the frequencies of input sound signals are by switching and selecting the output of a pitch extracting method through the use of a zero cross detection and the output of a pitch extracting method through the use of a digital filtering in accordance with the pitch frequency. CONSTITUTION:The note.number.pit value, which correspond to the pitch obtained by a zero cross detection method, and the prescribed musical scale are compared and a pronunciation processing is executed against a musical pronunciation circuit based on the processing results of the zero cross detection method when the input sound signal pitch is low and conversely when the input sound signal pitch is high, the processing is executed based on the processing results of the filter processing method. That is to say, a DSP 116 performs the zero cross detection or the filter process using a ROM 117 which stores a conversion table that converts from the various coefficients for digital filter operation based on the sound signals and pitch period to the note.number and a work RAM 18 which stores various data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pitch extracting apparatus for extracting a pitch (scale) of a waveform from an input acoustic signal, and an electronic musical instrument using the same.

[0002]

2. Description of the Related Art Conventionally, there has been a demand for extracting a pitch frequency or a pitch period (hereinafter simply referred to as "pitch") from an audio signal having a repetitive waveform such as a singing voice sung by a person or a musical tone collected from a microphone. Is strong. If the pitch can be extracted, the extracted pitch information can be used for various processes, for example, a musical tone corresponding to the pitch can be generated from an electronic musical instrument or the like.

Conventionally, various techniques have been proposed as techniques for extracting a pitch from an audio signal. As a first representative conventional example of the pitch extraction method, for example, as disclosed in Japanese Patent Application No. 3-120291 by the present applicant, adjacent zero-cross points (time positions where the amplitude becomes 0) of an input audio signal. Is determined as the pitch period of the waveform.

As a second representative conventional example of the pitch extraction system, for example, as disclosed by the present applicant in Japanese Patent Application No. Hei 2-123789, band pass filters corresponding to respective musical scales are provided, and each filter is provided with a filter. In this method, the level of the frequency spectrum related to the frequency corresponding to each scale is detected, and the frequency having the maximum level is determined as the pitch frequency.

[0005]

However, in the first conventional example, when the pitch frequency of the input acoustic signal becomes high, the pitch period becomes too short, and it is difficult to obtain an accurate pitch period. There is a problem that.

Further, the second conventional example has a problem that when it is constituted by an analog circuit, a plurality of band-pass filters must be provided, and the circuit scale becomes large. .

On the other hand, if a plurality of bandpass filters are realized by digital filters operating in a time-division manner, the circuit scale can be reduced. However, when the pitch frequency of the input acoustic signal decreases, the value of the filter coefficient increases because a low-band bandpass filter must be configured, and the number of bits required to express the filter coefficient becomes digital. For example, a DSP (Digital Signal Processor) that realizes a filter
The number of bits exceeds the allowable bit number.

When the second conventional example of the pitch extracting device is used as a musical scale detecting device, the difference between the scale frequencies in the high scale region is, for example, A6 = 1760 [Hz] and A
# 6 = 1867 [Hz] difference of 107 [Hz]
Although the frequency difference is relatively large, the difference between the scale frequencies in the lower scale region is, for example, A2 = 110 [Hz] and A # 2 = 116.
There is only a slight frequency difference, such as 7 [Hz] which is the difference between [Hz]. Therefore, a band-pass filter having a sharper cut-off frequency must be formed in a lower scale region, and as a result, the order of the band-pass filter becomes higher and D
There is a problem that real-time processing by SP or the like becomes impossible.

An object of the present invention is to make it possible to accurately detect any pitch frequency of an input acoustic signal, regardless of the pitch frequency.

[0010]

According to the present invention, first, a zero-cross point at which a value changes from a negative or positive first polarity to another second polarity is sequentially detected from an input acoustic signal. There is provided a first pitch detecting means for detecting pitch information of the audio signal from an interval between the zero cross points. The means sequentially obtains, for example, a zero-crossing interval from a zero-crossing point at which the polarity of the sound signal changes in a predetermined direction to a zero-crossing point at which the next polarity changes in the opposite direction, and one zero-crossing when the sequentially obtained zero-crossing intervals match. The pitch period of the sound signal is detected from the interval from the leading zero-cross point of the interval to the leading zero-cross point of the other zero-cross interval.

Next, digital filtering processes having different characteristics are sequentially performed on the audio signal in a time-division manner, whereby the level of each frequency spectrum for a plurality of predetermined frequencies corresponding to the process is detected. A second pitch detecting unit that detects pitch information of the acoustic signal by determining a level of the spectrum; The same means, for example, sequentially executes a high-pass filtering process of a predetermined characteristic and each low-pass filtering process to which the resonance having a peak at each of the predetermined frequencies is added as a time-division cascade process on the audio signal, Further, signal processing such as low-pass filter processing for extracting each envelope component from each output signal corresponding to each predetermined frequency obtained as a result is sequentially performed in a time-division manner, and each signal corresponding to each predetermined frequency obtained thereby is obtained. The level of each frequency spectrum described above is detected as an envelope component. Then, for example, the pitch information is detected as a frequency corresponding to the level having the maximum value among the levels of these frequency spectra.

Further, if the pitch information detected by any of the first and second pitch detecting means corresponds to a pitch frequency equal to or lower than a predetermined frequency threshold, the pitch information from the first pitch detecting means is used. Output means for outputting pitch information from the second pitch detection means if the pitch frequency corresponds to a pitch frequency equal to or higher than a predetermined frequency threshold.

According to the present invention, there is provided an electronic musical instrument having a tone generating means for generating a musical tone of a musical scale corresponding to the pitch information by using the pitch information from the pitch extracting apparatus having the above configuration. Can also.

[0014]

As described in "Problems to be Solved by the Invention", the pitch extraction method based on zero-cross detection is as follows.
Good performance is exhibited in a range where the pitch frequency of the input acoustic signal is low. On the other hand, the pitch extraction method using digital filtering exhibits good performance in a range where the pitch frequency of the input audio signal is high.

Therefore, in the present invention, high-precision pitch extraction utilizing both advantages can be realized by switching and selecting the output of these two systems according to the pitch frequency.

In this case, in particular, since the pitch extraction method by digital filtering is used only in a high pitch frequency range, the value of the filter coefficient can be reduced, and it is easy to use a normal DSP (digital signal processor) or the like. A digital filter can be configured as follows.

When the pitch extracting device according to the present invention is used as a musical scale detecting device of an electronic musical instrument, the pitch frequency is high, that is, the difference between the musical scale frequencies in the high musical scale range is a large value. There is no need to construct a rapid digital filter. As a result, the order of the bandpass filter can be reduced, and real-time processing by a DSP or the like can be easily realized.

[0018]

Embodiments of the present invention will be described below in detail with reference to the drawings. Overall Configuration FIG. 1 is a block diagram showing the overall configuration of an embodiment of the present invention.

In this embodiment, a CPU (Central Processing Controller) is used.
101, ROM 102, RAM 103, scale detector 11
1, a keyboard 106, a display 105, a printer 104, a tone generation circuit 108, a D / A converter 109, an amplifier 110, a speaker 111, and the like. These components are connected by a bus 107.

The ROM 102 stores various control programs shown in an operation flowchart described later and various data such as timbre data. The CPU 101 uses the RO
The other parts are controlled according to the control program in M102.

The RAM 103 stores a scale detection unit 11 described later.
1 stores data and the like for operating 1 as a zero-cross detection device, a digital filter, or an envelope extraction device, and also functions as a work memory of the CPU 101.

The scale detecting section 111 includes a microphone 112, a microphone amplifier 113, a low-pass filter 114, an A / D converter 115, a DSP 116, a ROM 117, and a work RA.
M118, and has LINE IN as a line input terminal for audio signals.

The scale detecting section 111 includes a microphone 112
Alternatively, an analog sound signal (instrument sound, human voice sound, or reproduction sound from a tape recorder, a radio, a television, a CD player, or the like) input from a line input terminal LINE IN is input to an A / D converter via a low-pass filter 114. The signal is converted to a digital audio signal (hereinafter simply referred to as an audio signal) x (n) and input to the DSP 116.

The DSP 116 is based on the audio signal x (n), and is a ROM 117 storing a conversion table for converting various coefficients and a pitch period into a note number for a digital filter operation to be described later, or a work for storing various data. Using the RAM 118, zero-cross detection or filter processing described later is performed.

The processing result of the DSP 116 is transmitted to the CPU 10
Sent to 1. The CPU 101 generates the tone generation circuit 1 based on the processing result of the DSP 116 and the storage states of various modules incorporated in the tone generation circuit 108 described later.
Assignment processing and deletion processing of the scale sound to 08 are performed.

The keyboard 106 is provided with function switches, a keyboard, and the like. When these are operated, the CPU 101 detects the operation state and assigns a tone to be generated to the tone generator 108.

Display 105 or Printer 104
Operates under the control of the CPU 101 to display the detected scale or to print the scale on paper. For example, the CPU 101 displays the scale of the audio signal being input in real time on the display 1.
The musical scale of the audio signal may be displayed on the display 105 as a musical score or printed on paper by the printer 104 after editing work or the like in non-real time.

As the tone generating circuit 108, for example, P
Various types of sound source circuits such as a CM system, an FM system, an iPD system, and a sine wave synthesis system can be applied. The tone generation circuit 108 has a plurality of (for example, four) tone generation channels, and generates a scale tone having a scale number assigned by the CPU 101. This scale sound is emitted from the speaker 110 via the audio system 109. FIG. 2 is an overall configuration diagram of the DSP 116 shown in FIG.

Referring to FIG.
1 is a bus 107 connected to the CPU 101 and an A / D
The buses connected to the conversion 115 are accommodated, and each bus and the DSP
Connect to internal circuit.

The operation ROM 203 stores the DSP 1
16 is a ROM that stores a microprogram that defines the entire operation of the microcomputer 16, and reads a corresponding program instruction based on a designated address from the address counter 202. The CPU 101 in FIG.
By setting data in
It instructs the address counter 202 what program to read from the OM 203 and execute each process described below.

The output of the operation ROM 203 is also supplied to a decoder 204, which outputs various control signals to each circuit in the DSP 116 to perform a desired operation. On the other hand, the internal bus of the DSP 116 has a ROM 117 of FIG.
And the work RAM 118 are connected. Then, according to the program command from the operation ROM 203,
A filter coefficient, an acoustic signal x (n), and the like are supplied to the DSP 116 or input / output to / from the work RAM 118.

The register group 208 is composed of a plurality of registers for temporarily storing data being operated on, and is connected to each input / output terminal of the multiplier 205 or the adder / subtractor 207 via an internal bus. Then, in order to implement a judging process based on the operation result (comparison result or the like) from the adder / subtractor 207, the address counter 2 via the flag register 206 is used.
To 02, a flag signal indicating the judgment result is transmitted.

The address of the address counter 202 is changed according to the output of the flag register 206, and the program instruction is read from the operation ROM 203 according to the address. In this way, a judge process is realized.

The operation of the embodiment of the present invention having the configuration shown in FIGS. 1 and 2 will be described below in order. Basic Operation of DSP 116 and CPU 101 First, the overall operation of the DSP 116 will be described.

FIG. 3 shows the operation ROM 20 of FIG.
6 is a flowchart illustrating an overall operation of the DSP 116 realized by executing a microprogram stored in the DSP 3.

First, when the power is turned on, all contents such as the work RAM 118 and the register group 208 in FIG. 2 are initialized (step S301). Next, the A / D converter 115 of FIG. 1 waits for the A / D conversion processing for one sample to be completed (step S302). When the A / D conversion processing is completed, the A / D converter 1 converts the data into digital data. The sound signals for the samples are taken into the work RAM 118.

Subsequently, based on the A / D-converted sound signal, pitch detection is performed by zero-cross detection described later (step S303). Further, based on the A / D-converted sound signal, pitch detection is performed by a filtering process described later in detail (step S30).
Four).

The work RA is detected by the zero-cross detection.
Note number corresponding to the pitch obtained in M118
The PIT, the envelope value ENV, and the envelope signal Et (n) of each scale also obtained in the work RAM 118 by the filtering process are transferred to the CPU 101 (step S101).
S305).

The processing in steps S302 to S305 is performed by A /
It is executed each time the A / D conversion processing for one sample is completed in the D converter 115. As described above, DSP11
6 is characterized in that pitch detection by zero-cross detection and pitch detection by filter processing are respectively performed on an acoustic signal.

Next, the overall operation of the CPU 101 will be described. FIG. 4 shows a CP realized by executing a control program stored in the ROM 102 of FIG.
It is a flowchart which shows the whole operation | movement of U101.

First, when the power is turned on, the RAM shown in FIG.
118, the contents of a register (not shown) in the CPU 101 are initialized (step S401). Next, the DSP
It is awaited that data based on the transfer processing in step S305 in FIG. 3 is transferred from step 116 (step S40).
2) When the data is transferred, the DSP 11
It is determined whether the value of the note number PIT corresponding to the pitch obtained by the zero-cross detection in step 6 (step S303 in FIG. 3) is below a predetermined scale K (step S303).
S403).

Then, the determination in step S403 is YES,
If it is determined that the pitch of the input acoustic signal is low, D
Zero cross detection in SP116 (step S303 in FIG. 3)
Based on the processing result, the tone generation circuit 108 performs a tone generation process (step S404).

Conversely, if the determination in step S403 is NO and it is determined that the pitch of the input acoustic signal is high, the DSP
Based on the processing result of the filter processing at 116 (step S304 in FIG. 3), tone generation processing is executed on the musical tone generation circuit 108 (step S405).

The processing of steps S402 to S405 is performed by the DS
It is executed every time data based on the transfer processing in step S305 of FIG. 3 is transferred from P116. Zero Cross Detection Processing Next, zero cross detection processing in the DSP processing of FIG. 3 will be described. <Basic Operating Principle of Zero-Cross Detection> Normally, a voice that has a clear pitch feeling and is played as a single sound, such as a human singing voice, has a pitch period (for example) as shown in FIG.
A substantially similar waveform element repeated every time (for example, a
And d, b and e, and c and f).

Such a sound has many zero-cross points in one pitch cycle whose waveform crosses a reference level having an amplitude of 0 due to harmonic components contained in the sound.
Therefore, in order to determine the pitch period from such a periodic waveform, it is necessary to determine which of the zero-cross points should be the pitch period. It has the following features.

That is, the intervals (for example,) defined by the zero cross points (for example, Z1 to Z2 and Z7 to Z8) of the substantially identical waveforms (for example, a and d) of adjacent pitch periods are approximately equal.

In this embodiment, focusing on the above-described points, finding two sections (for example, and) having substantially the same time between adjacent zero-cross points, the start points (for example, Z1 and Z7) of the two sections are determined. ) Is extracted to detect a note number corresponding to the pitch of the input audio signal.

The operation of zero-cross detection that operates according to the above-described basic operation principle will be described below. <Outline of Zero-Cross Detection Operation> In FIG. 5, for example, the time of the section of the zero-cross points Z1 to Z2 of the waveform a is counted, and the count value COUNT 1 is held in the first counter. Next, a waveform b having the same polarity as the waveform a,
.. are counted sequentially, and the count value COUNT 2 is sequentially updated until a value substantially equal to COUNT 1 is obtained.
Is held in the counter.

In this case, the time from the zero-cross point Z1 to the zero-cross points Z3, Z5 and Z7 of the waveforms b, c and d is counted by the third counter. Since the count value COUNT 3 has a possibility of being the pitch period of the input audio signal, R is sequentially updated while a count value corresponding to the pitch period is obtained.
Stored in the variable PITCOUNTER in AM4.

When a section having a time substantially equal to the time of the section is found, the counter value corresponding to the time between the zero-cross points Z1 and Z7 is stored in the memory variable PIT COUN.
It is read from the TER and determined as the pitch period. The note number corresponding to the pitch cycle is R
It is obtained from the table in the OM 117. <Specific Operation of Zero-Cross Detection> Next, the operation for obtaining the pitch of the waveform of the input acoustic signal according to the above-described operation principle will be described with reference to the operation flowcharts of FIGS. These operation flowcharts correspond to step S in FIG.
Details of a part 303 are realized by the DSP 116 as an operation of executing a microprogram stored in the operation ROM 203.

In FIG. 6, first, the work RAM 118
Includes a status flag (see FIG. 5) indicating the state of the signal waveform at the time of reading the audio signal (see FIG. 5), a variable ENV indicating the magnitude of the waveform envelope of the input audio signal, and a pitch extraction A variable PIT holding a note number corresponding to the resulting pitch, a variable count 1, count 2, count 3, and 1 holding a count value of each of the first, second, and third counters. Each data such as a variable ADMEM which holds the previous A / D conversion value is held.

Subsequently, one sample of the audio signal input from the A / D converter 115 is stored in the register reg1 in the register group 208 (step S601). Thereafter, the process branches into three depending on the status flag (step S602). Initially, since the status flag of FIG. 5 is 0, the process proceeds to step S603. In the state where the status flag = 0, the amplitude of the input audio signal is detected from the negative side to the positive side to detect a zero crossing point Z1.
It is determined whether the value of the variable ADMEM in which the previous A / D conversion value is written is negative and the value of the register reg1 in which the current A / D conversion value is stored is positive.

In this case, the determination in step S603 is NO, and the process proceeds to step S605. Then, the value of reg1 in which the A / D conversion value is set this time is moved to the variable ADMEM (step
After the envelope extraction (step S606) is performed, the zero-cross detection ends.

The process of the envelope extraction (step S606) is shown as an operation flowchart of FIG. In FIG. 8, first, the current A / D conversion value (contents of reg1)
Is calculated (step S801), and the absolute value data is passed through a low-pass filter to shape the waveform of the input acoustic signal (step S802). Then, the resulting envelope value is stored in the variable ENV (step S803).

In the above processing, each sample of the audio signal is input, and the zero-cross detection in step S303 of FIG. 3 is repeatedly executed, and the sampling point of the input audio signal is changed to the zero-cross point (the amplitude of which goes from the negative side to the positive side). This is repeated until Z1) in FIG. 5 is reached.

When the sampling point reaches the zero cross point, the determination in step S603 in FIG. 6 becomes YES, and the status flag STATUS FLAG is set to 1 (step S603).
S604).

As described above, the status flag STATUS F
Figure 3 corresponding to the next sample point because LAG is 1
In the process of step S602 in the zero cross detection of S303, the process proceeds to step S607.

In this state, the increment of the variable COUNT 1 (hereinafter, referred to as the first counter COUNT 1) as the first counter is counted in order to count the time exemplified in the section of FIG. Variable COUNT 3 as a third counter to count the time exemplified in
(Hereinafter, referred to as a third counter COUNT 3) is started.

First, the value of the first counter COUNT 1 is incremented by 1 (step S607). Next, the zero cross point where the amplitude of the input acoustic signal goes from the positive side to the negative side, that is, FIG.
In order to detect the zero crossing point Z2, which is the end point of the section, the value of the variable ADMEM in which the previous A / D conversion value is written is positive and the register in which the current A / D conversion value is written It is determined whether the value of reg 1 is negative (step S608).

If the determination is NO, the third counter COUN
The value of T 3 is increased by 1 (step S610). Since then
Step S608 → S610 → S605 → S606 ~ S602 → S607 → S608 ・
The loop processing is repeated every time each sample of the audio signal is input and the zero-crossing detection is performed in step S303 in FIG. 3, and the first counter CO
The values of UNT 1 and the third counter COUNT 3 are incremented by one.

Then, if the determination in step S608 is YES and the sampling point of the input audio signal reaches the zero crossing point Z2 where the amplitude goes from the positive side to the negative side, the status flag is set to 2 (step S609). . Thereafter, the process proceeds to step S610, and after COUNT 3 is increased by 1, the process proceeds from step S605 to S606.

After the above-described loop processing is performed, FIG.
The counter value corresponding to the time of the section is the first counter
Stored in COUNT 1. Subsequently, when the next sample of the acoustic signal is input and the zero-cross detection is performed again in step S303 in FIG.
Since the flag is set to 2, the process proceeds from step S602 to step S611 → S612 in FIG. 7 this time.

In step S612, the same processing as step S603 in FIG. 6, ie, the variable AD, is performed to detect the zero cross point (Z3 in FIG. 5) where the amplitude of the input audio signal goes from the negative side to the positive side.
It is determined whether the value of MEM (the previous A / D conversion value) is negative and the value of the register reg1 (current A / D conversion value) is positive.

While the determination is NO, S612 → step S610 (FIG. 6) → S605 → S606-S602 → S611 (FIG. 7) → S6
The loop processing of 12... Is repeated every time each sample of the audio signal is input and the zero-cross detection in step S303 of FIG. 3 is executed.
The count value of 3 is incremented by one (S610).

Then, when the determination in step S612 is YES and the sampling point of the input audio signal reaches the zero-crossing point Z3 whose amplitude goes from the negative side to the positive side, the status flag is set to 3 and the status flag at this time is set. COUNT 3 value (Figure 5
Is stored in the variable PITCOUNTER (step S613). The value of the PIT COUNTER is held as a candidate for the cycle corresponding to the pitch to be extracted, that is, as a value that has a possibility of becoming the count value of the section in FIG.

Thereafter, steps S613 → S610 (FIG. 6) → S6
After proceeding from 05 to S606, the next sample of the audio signal is input, zero-cross detection is performed again in step S303 in FIG. 3, and the process proceeds to S601 to S602 to S611 (FIG. 7).

This time, since STATUS FLAG = 3, the process proceeds from step S611 to step S614, where the count value of the second counter COUNT 2 is incremented by 1, and from this point on, the counting of the divisions in FIG. 5 is started. .

Next, a zero cross point where the amplitude of the input audio signal goes from the positive side to the negative side, that is, a zero cross point Z4 corresponding to the end point of the section in FIG. 5, is detected.
(The previous A / D conversion value) is positive and the register reg1
It is determined whether or not the value (current A / D conversion value) is negative (step S615).

While the determination is NO, steps S614 → S6
15 → S610 (Fig. 6)-S606-S601 → S602 → S611 (Fig. 7) →
The loop processing of S614 is repeated each time each sample of the audio signal is input and the zero-cross detection of step S303 in FIG. 3 is executed, and the value of the second counter COUNT 2 and the value of the third counter COUNT 3 are repeated each time. Is incremented by one.

If the determination in step S615 is YES, it means that the zero cross point Z4 has been detected, and the status flag is changed to 4 here (step S616). Thereafter, the process proceeds to step S610 in FIG. 6, and after COUNT 3 is increased by 1, the process proceeds from step S605 to S606, the next sample of the acoustic signal is input, and the zero cross detection in step S303 in FIG. 3 is executed again. Then, the process proceeds from S601 to S602 to S611 (FIG. 7).

This time, since STATUS FLAG = 4, the process proceeds from step S611 to step S617. In this state, the divisions are sequentially compared with the divisions in FIG. 5, and when a division having a time substantially equal to the division is found, a processing operation for determining the division time as the pitch period of the input acoustic signal. Is performed.

First, the first counter CO obtained so far is
It is determined whether or not the respective values of UNT 1 and the second counter COUNT 2 are substantially equal, that is, whether or not the absolute value of the difference between the respective values of COUNT 1 and COUNT 2 is within an allowable error range. (Step S617). This allowable error range depends on the sampling frequency and pitch frequency of the input audio signal, but is empirically determined.

In this case, since COUNT 1 and COUNT 2 each hold the count value of the section in FIG. 5, the determination in step S617 is NO, and the process proceeds to step S618. Here, if noise or the like is mixed in the input audio signal, the state where the determination in step S617 is NO may be abnormally long, and the tone generation instruction may not be performed.

In order to prevent the occurrence of such an abnormal state,
The value of a third counter COUNT 3 that counts the time to be taken as the pitch period of the sound signal is constantly monitored, and the value of the third counter COUNT 3 is a predetermined limit value having a value much larger than the pitch period of the lowest sound of the sound signal. Is determined (step S618).

This limit value is also similar to the above-described allowable error range.
Determined empirically. For example, set the sampling frequency to 20KH
If the minimum extraction pitch (pitch) is 110.5 Hz (corresponding to the pitch name: A1), then 20000 ÷ 110.5 = 180.995..., the limit value may be set to around 180.

If the determination in step S618 is YES, that is, if the state is abnormal, an instruction is given to the tone generation circuit 108 via the CPU 101 in FIG. R variable on now is cleared, and the process of zero-cross detection ends.

If the determination in step S618 is NO, the process proceeds to step S619, where the value of the second counter COUNT 2 is cleared, and the status flag is returned to 2. And
The process for obtaining the next zero cross again is started. At this time, the value held in the variable PIT COUNTER becomes invalid, but the third counter COUNT 3 is still counting up, and the first counter COUNT 1 also holds the count value in the section of FIG. ing.

As described above, the status flag 2
After returning to the state, the status flag advances to the state of 3 and 4 as in the previous case, and the second counter COUNT
2 holds the count value of the section in FIG. 5, and the variable PIT COUNTER holds the count value (of FIG. 5) from the start of the section to the zero crossing point z5 of the section.

However, the value of COUNT 1 (corresponding to the time of the section) and the value of COUNT 2 (corresponding to the time of the section) do not match this time, so that the determination in step S407 is NO. Therefore, the process proceeds to step S619 via step S618, and the status flag is again executed from the state of 2.

Then, when the status flag becomes the state of 4 again, COUNT 2 holds the count value of the section of FIG. 5, while PIT COUNTER holds the count value of the time of FIG. ing. Then, in step S617, YES is determined for the first time, and the process proceeds to step S621.

The correct value of the pitch period of the input audio signal at this time is the value of the third counter COUNT 3 (FIG. 5) held in PIT COUNTER. The note number corresponding to this value is stored in the ROM 117 of FIG.
Is read out from the conversion table and moved to the variable PIT (step S621).

Next, the status flag and all counters are cleared and reset to the initial state (step S622). Then, after proceeding from step S605 to S606 in FIG. 6, the process of zero-cross detection ends.

As described above, each time the sample of the acoustic signal is input and the zero-cross detection is performed in step S303 in FIG. 3, the note number PI corresponding to the pitch period is obtained.
T and the envelope value ENV are obtained, and these are transmitted from the DSP 116 to the CPU 10 by the processing of step S305 in FIG.
1 and is used in the sound generation process of step S404 in FIG.

In the above-described zero-crossing detection processing, note numbers (pitch) are extracted by focusing only on the regularity of the zero-crossing interval. If regularities such as the area and the maximum value of the amplitude are taken into account, higher-precision processing becomes possible. Next, the filter processing in the DSP processing of FIG. 3 will be described. <Basic Operation Principle of Filter Processing> First, the basic operation principle of the filter processing executed by the DSP 116 will be described with reference to FIGS.
Explanation will be made using 0.

FIG. 9 is a basic functional block diagram of the filtering process. In FIG. 9, the input audio signal x (n) is supplied from the A / D converter 115 of FIG.
This is an audio signal for one sample input to the SP 116. Note that n indicates a discrete sampling time.

For the above-mentioned input audio signal x (n), DS
By the time division processing in P116, filtering processing of N bandpass filters is executed. now,
Let Ht (z) be the transfer function of a bandpass filter of an arbitrary scale t (1 ≦ t ≦ N). In this embodiment, each transfer function Ht
(z) The characteristic of (1 ≦ t ≦ N) is changed depending on each scale of a plurality of octaves. FIG. 10 shows a band-pass filter having a transfer function Ht (z) (hereinafter, band-pass filter H
FIG. 11 is a frequency characteristic diagram in the case where a Chebyshev filter is adopted as (t (z)). The transfer function Ht in this case
(z) is represented by the following equation.

[0087]

(Equation 1)

Here, if i = 1, the DSP 116
Is

[0089]

(Equation 2)

The calculation of the digital filtering process shown in FIG. Further, when i ≧ 2, the same operation as in Expression 2 is repeatedly executed.

The filter coefficients of the band-pass filter Ht (z) can be obtained by numerical calculation. As a specific example, the center frequency corresponding to note number A4 is 44
When the band-pass filter of 0 [Hz] is configured under the following conditions (see FIG. 10), digital filtering processing having a transfer function Ht (z) realized by the following coefficient values is executed. Will be.

That is, under the following conditions: = 1 [dB] = Sampling frequency fs = 10 [kHz] = 12 [dB] or more = 415 [Hz] = 430 [Hz] = 450 [Hz] = 466 [Hz] The filter coefficients of the two-stage digital filter of = 1, 2 are as follows: For Ht (0) = 0.08192384 i = 1, a ₁ (1) =-1.91442200776 a ₂ (1) = 0.9933673 b ₁ (1) = -1.91105345727 b ₂ (1) = 1. For i = 2, a ₁ (2) =-1.9210712 a ₂ (2) = 0.993606 b ₁ (2) =-1.93525314797 b ₂ (2) = 1. .

As described above, the digital filtering process of the band-pass filter Ht (z) of each scale t (1 ≦ t ≦ N) is executed by the DSP 116 in a time-division manner. As a result, each output signal Yt ( n) (1 ≦ t ≦ N) is obtained.

Subsequently, as shown in FIG. 9, the DSP 116 performs an envelope extraction process on each of the output signals Yt (n) by a time division process. This envelope extraction process is performed for each output signal Yt (n).
This is realized as a process of obtaining a peak level (absolute value) at predetermined time intervals (for example, time intervals corresponding to each musical scale). Alternatively, it is realized as a process of executing a specific digital filtering process as described later on the absolute value of each output signal Yt (n).

As described above, the envelope signal Et (n) for each scale t (1 ≦ t ≦ N) is obtained by the time division processing by the DSP 116. And, C
In the sound generation process of step S405 in FIG. 4 by the PU 101, the scale information can be detected from the input audio signal x (n) by executing a level determination process as described later in detail. <Improved operation principle of filter processing> The above-described band-pass filter Ht (z) has been described in connection with the case where it is realized as a Chebyshev-type filter. For example, the center corresponding to the note number A4 as described above is used. When a bandpass filter having a frequency of 440 [Hz] is configured, the digital filtering requires eight multiplications at each sample timing.

The improved operation principle of the filtering process that reduces the number of multiplications in the digital filtering process and enables real-time filtering will be described below with reference to FIGS.

FIG. 11 is an improved functional block diagram of the filter processing. In FIG. 11, the transfer function Ht (z) (1 ≦
t ≦ N), a high-pass filter process having a transfer function H1 (z) (common to all scales t) and a transfer function H2t (z) (1 ≦ t)
≦ N) by cascade processing with low-pass filter processing. In this case, the transfer function Ht (z) of the band-pass filter in FIG. 9 is represented by the product of the transfer function H1 (z) of the high-pass filter and the transfer function H2t (z) of the low-pass filter as shown in FIG.

Here, the transfer function H1 of the high-pass filter
(z), which will be described in detail later, is 0 at the frequency 0, and the frequency fs /
It has the characteristic that the output becomes maximum at 2 [Hz]. The transfer function H2t (z) of the low-pass filter has a resonance-type characteristic in which the output becomes maximum at each scale frequency, which will be described in detail later. Therefore, the transfer function Ht (z) of the filter obtained by cascading the above-described high-pass filter H1 (z) and low-pass filter H2t (z).
Are characteristics of a pseudo bandpass filter as shown in FIG. In FIG. 12, each center frequency f _{1 to} f _N corresponds to each scale frequency, and the frequency difference between adjacent bands is Δf
It is. Furthermore, the value of N can be set to about 40 to 50 (in the case of this embodiment, 3 to 4 octaves on the high frequency side).

Each output signal Wt (n) of the low-pass filter H2t (z) obtained by such a filtering process.
(1 ≦ t ≦ N) is the transfer function H _Et (z) as shown in FIG.
An envelope signal Et (n) (1 ≦ t ≦ N) is obtained as an output by inputting to each of the envelope extraction units represented by (1 ≦ t ≦ N). This envelope extraction unit is realized as a low-pass filter having a low cutoff frequency, as described later in detail.

Next, the filter characteristics of the high-pass filter having the transfer function H1 (z), the low-pass filter having the transfer function H2t (z), and the envelope extracting section having the transfer function H _Et (z) will be described in detail below. Will be described.

FIG. 13 shows the high-pass filter H1 of FIG.
FIG. 2 is a configuration diagram showing (z) by a hardware image.
This is a second-order FIR digital filter whose transfer function is

[0102]

(Equation 3)

Are shown. In FIG. 13, 1301,
Z ^{-1 in} 1302 represents a delay element for delaying one cycle of the sampling clock, and 1306 and 1307 are adders. Reference numerals 1303, 1304, and 1305 denote multipliers. Each of the coefficients x2, x1, and x1 / 4 is a multiplication coefficient in each of the multipliers.

In the DSP 116, filter processing equivalent to the high-pass filter in FIG.

[0105]

(Equation 4)

This is realized by the following discrete operation processing. In this case, since the filter coefficient is a multiple of 2, the multiplication of the coefficient and the signal is realized by simple bit shift processing.

The high-pass filter has the following frequency characteristics:

[0108]

(Equation 5)

The characteristic is such that the gain is minimum when Ω = 0 (0 [Hz]) and maximum when Ω = π (fs / 2 [Hz]). Here, fs is the A / D converter 11 of FIG.
5 is the sampling frequency of the acoustic signal x (n). FIG. 14 shows the characteristics of this high-pass filter. According to the sampling theorem, the frequency characteristics from fs / 2 to fs [Hz] are such that the frequency characteristics from 0 to fs / 2 [Hz] are turned around at fs / 2 [Hz].

FIG. 15 is a block diagram showing the hardware of the low-pass filter H2t (z) shown in FIG. This is a second-order IIR digital filter whose transfer function is

[0111]

(Equation 6)

## EQU12 ## Then, in this equation, θ and CY
Varies according to each scale t, and r
Is a parameter indicating the intensity of resonance, that is, the magnitude of the output peak.

In FIG. 15, 1501 and 1502 are
Representing delay elements for delaying one cycle of the sampling clock, 1503, 1504 and 1505 are multipliers,
Coefficient shown in the drawing, each -2rcosθ, r ^2,
CY is multiplied. 1506 and 1507 are adders.

In the DSP 116, a filter operation equivalent to the low-pass filter having the configuration shown in FIG.

[0115]

(Equation 7)

[0116]

(Equation 8)

This is realized by the following discrete operation processing. Here, the pole of the transfer function is z ₁ with “**” as a power operation.
= Re ** (jθ), z ₂ = re ** (-jθ), and z =
There is a double zero at zero. The arrangement of the poles and zeros and 0 <θ
FIG. 16 shows the relationship between the pole vector and the zero point vector when <π / 2 is set.

As can be understood from the figure, assuming that the angle formed by the line segment connecting the point P moving along the unit circle and the zero point 0 with the real axis is Ω, the pole vector Z ₁ and Ω = 0 to π the difference is that the length of the vector V ₂ of the zero point vector V ₁ was initially reduced and then increases to. The minimum length of the vector V ₂ is when the point P is closest to the pole Z ₁ , that is, θ
= Ω.

Here, the frequency Ω of this low-pass filter
Magnitude of the frequency response (amplitude characteristics) of is determined by the respective ratio of the length of the zero vector V ₁ and the vector V _2. Since the value of the zero point vector V ₁ is always 1, the magnitude of the frequency response is proportional to the reciprocal of the magnitude of the vector V ₂ , and becomes maximum at θ = Ω as described above. The peak of the magnitude of the frequency response is determined by the value of r,
It increases as the value of r approaches 1. FIG. 17 shows that Ω = −
The magnitude of this frequency response in π to π is shown.

On the other hand, the phase of the frequency response is a value obtained by subtracting the angle between the real axis and the vector V ₂ from the angle Ω between the real axis and the zero point vector V ₁ . As apparent from the above description, with reference to the center frequency f _t and the sampling frequency f _s of the scale t (1 ≦ t ≦ N) , it is determined the value of theta obtained at θ = 2πf _s / f _t, FIG. 18 as shown in, a low-pass filter having a peak at each band of the center frequency f _t can be realized.

In this case, the magnitude of the peak varies depending on the value of r as described above. The value of r is selected to prevent the peak from affecting the adjacent band, and the value of CY described above is selected. Can be set so that the level of the output Wt (n) of each band becomes substantially equal.

The values of r and CY can be obtained, for example, as follows. Now, the center frequency f _t of each scale _t
And the center frequency f of the next scale t + 1 that is Δf away from it
The ratio of the magnitude of the frequency response to _{t + 1} (= f _t + Δf) is represented by m
Then

[0123]

(Equation 9)

By solving a quartic equation of r, as a result, among the obtained r, r satisfying 0 <r <1 is selected, and each coefficient −2rcos θ, r ² can be obtained. Numerical results of the calculations, for example the center frequency is 440 corresponding to the note number A4 [Hz], the sampling frequency f _s = 5 [kHz], When m = 4, -2rcosθ = -1.9773 r 2 = 0.9851 CY = 36.7 Becomes The other bands can be obtained in the same manner.

The low-pass filter having the transfer function H1 (z) and the low-pass filter having the transfer function H2t (z) are connected in cascade as shown in FIG. As described above with reference to FIG. 12, the characteristics of the pseudo bandpass filter are realized by the overall transfer function Ht (z) expressed as

FIG. 19 is a block diagram showing a hardware image of the low-pass filter transfer characteristic H _Et (z) corresponding to the envelope extraction unit in FIG. This is a second-order IIR digital filter having the same form as the low-pass filter H2t (z) described above, and its transfer function is

[0127]

(Equation 10)

Is as follows. This equation shows that in the transfer function H2t (z) of the low-pass filter described above in equation 6, r = 0.9,
It is equivalent to θ = 0. In FIG. 19, the absolute value circuit 1901 outputs the absolute value | Wt (n) | of the output Wt (n) of the low-pass filter H2t (z) in FIG. 11 and sends it to the next digital filter unit. 1902 and 1903 are delay elements for delaying one cycle of the sampling clock.
Reference numerals 4, 1905 and 1906 denote multipliers, each of which is multiplied by each coefficient shown in FIG. Reference numerals 1907 and 1908 denote adders.

In the DSP 116, a filter operation equivalent to the low-pass filter having the configuration shown in FIG.

[0130]

[Equation 11]

This is realized by the following discrete operation processing. As shown in FIG. 20, the frequency characteristic of this low-pass filter has a resonance having a maximum value at Ω = 0. The cut-off frequency of the low-pass filter is set to be much lower than the cut-off frequency of the low-pass filter H2t (z) in the lowest band for the purpose of extracting the envelope.

Here, the coefficient CE is for adjusting the output level of each musical scale t. FIG.
FIG. 20 is a diagram schematically showing the envelope signal Et (n) obtained by the low-pass filter of FIG. 19 in comparison with the absolute value | Wt (n) | of the output of the low-pass filter H2t (z). After the negative peak value (broken line in FIG. 21) of the input signal is converted into a positive peak value by an operation corresponding to the absolute value circuit 1901 in FIG. 19, low-pass filtering is performed. As a result, an operation for obtaining an envelope for each scale t is realized. <Specific operation of filter processing> An operation of obtaining a scale (pitch) of an input acoustic signal in accordance with the above-described improved principle of the filter processing will be described with reference to an operation flowchart of FIG. This operation flowchart corresponds to step S in FIG.
Details of a part 304 are realized by the DSP 116 as an operation of executing a microprogram stored in the operation ROM 203.

Each time the sample x (n) of the audio signal is input and the filter processing of step S304 of FIG. 3 is repeatedly executed by the filter processing according to the operation flowchart of FIG. 22, each scale of FIG. The processing corresponding to the high-pass filter section and the low-pass fill timing section for each scale and the envelope extraction section is executed by time-division processing, so that the envelope signal Et (n) of each scale is obtained for each sampling period. Is transferred to the CPU 101 by the process of step S305 in FIG.

In connection with the operation flow chart of FIG. 22, the work RAM 118 stores, in the work RAM 118, each variable x () which holds the current one-sample and two-sample previous sound signals input from the A / D converter 115 of FIG. n), x (n-1) and x (n-2)
, S (n), s (n−1) and s (n−2) holding the current output of one sample before and two samples before of the high-pass filter unit having the transfer function H1 (z) of FIG. ), The variables Wt (n) and Wt (n-1) that hold the current output of one sample before and two samples before of the low-pass filter corresponding to each scale t having the transfer function H2t (z) in FIG. ) And Wt (n-2)
(1 ≦ t ≦ N), and the transfer function H _Et (z) in FIG.
, The variables Et (n), Et (n-1) and Et (n-2) (1 ≦ 1) that hold the current output of one sample before and two samples before of the envelope extraction unit corresponding to each scale t having t ≦ N)
Are respectively stored.

In FIG. 22, first, the A / D converter 11
The audio signal for one sample input from the input unit 5 is shown in FIG.
Is stored in the variable x (n) in the work RAM 118 (step S2201).

Next, each variable x in the work RAM 118
From (n), x (n-1) and x (n-2), the register group 20 of FIG.
The acoustic signals of the current sample and the past two samples are read out to each register in the block 8, and a high-pass filter process represented by a transfer function H1 (z) in FIG.
S2202). This processing is an arithmetic processing represented by the above-described Expression 4, and is executed using the multiplier 205 and the adder / subtractor 207 in FIG. At this time, each filter coefficient used for the calculation of Expression 4 is read from the ROM 117. The resulting output is the variable s in work RAM 118
(n).

The above-described high-pass filter processing is common to each band, and is therefore performed only once. Next, a low-pass filtering corresponding to each chromatic t having a transfer function H2t in FIG 11 (z), the envelope extraction process corresponding to each chromatic t having a transfer function H _Et (z) are continuously executed.

These processes correspond to each scale t,
This is repeated as time division processing for the scale. for that reason,
A register t for repetition control for performing time division processing for N scales is provided in the register group 208 of FIG. 2, and the value of the register t is initialized to 1 in step S2203.

Each time the low-pass filter processing and envelope extraction processing for one musical scale are completed in steps S2204 and S2205, it is determined in step S2206 whether or not the content of the register t has reached N. , Step
In step S2207, the content of the register t is incremented, and the processes in steps S2204 and S2205 are repeated.

This judgment processing is performed by the adder / subtractor 20 shown in FIG.
7 and the flag register 206, and the address counter 202 determines in steps S2204 and S2205
Operation ROM is the program instruction group corresponding to
It is repeatedly read from 203.

First, the filter coefficients CY, 2rc of the scale corresponding to the number indicated by the register t are read from the ROM 117.
osθ and r ² are read. And work RAM
From the variables s (n), s (n-1) and s (n-2) of FIG.
The output signals of the high-pass filter processing in step S2202 for the current sample and the past two samples are read out to each register in the register group 208, and the signals shown in FIG. Function H
A low-pass filter process represented by 2t (z) is executed (step S2204). This processing is an arithmetic processing represented by the above-described Expression 7, and is executed using the multiplier 205 and the adder / subtractor 207 in FIG. The resulting output is stored in variable Wt (n) in work RAM 118.
Note that the subscript t of this variable corresponds to the current value of the register t.

Subsequently, from the ROM 117, the filter coefficient CE of the scale corresponding to the number indicated by the register t, "1.
8 ”and“ 0.81 ”are read out. Then, the contents of the variable Wt (n), which is the output of the above low-pass filter processing, the output Wt (n-1) of the low-pass filter processing of the past two samples read from the work RAM 118 to the registers in the register group 208, and An envelope extraction process (low-pass filter process) represented by the transfer function _Het (z) in FIG. 11 is performed on Wt (n-2) using the above-described filter coefficients (step S2205). This processing is an arithmetic processing represented by the above-described Expression 11, and is executed using the multiplier 205 and the adder / subtractor 207 in FIG. The output obtained as a result is stored in a variable Et (n) in the work RAM 118.
Note that the subscript t of this variable corresponds to the current value of the register t.

As described above, by repeating the processing of steps S2204 to S2207, the processing corresponding to the low-pass filter section and the envelope extracting section for N scales in FIG. 11 is performed.
t (n) is obtained, and is transferred to the CPU 101 by the processing in step S305 in FIG. Next, the sound generation processing by the zero-cross detection processing in the CPU processing of FIG. 4 will be described.

As described above, in the determination in step S403 in FIG. 4, the value of the note number PIT obtained by the zero-cross detection by the DSP 116 (step S303 in FIG. 3) is lower than the predetermined scale K, that is, If it is determined that the pitch of the input audio signal is low, in step S404, a tone generation process is executed on the tone generation circuit 108 based on the processing result of the zero-cross detection by the DSP 116.

FIG. 23 is an operation flowchart of the sound generation process by the zero-cross detection process of step S404 in FIG. This operation flowchart shows that the CPU 101
This is realized as an operation of executing the control program in M103.

In connection with the operation flowchart of FIG.
The RAM 103 holds a variable PIT for storing a note number corresponding to the pitch transferred from the DSP 116, a variable ENV for storing an envelope value, a variable ON NOW for storing a currently sounding note number, and the like. .

In FIG. 23, first, the value of the envelope value ENV transferred from the DSP 116 exceeds a predetermined threshold value THRESHOLD, and the value of the note number PIT is 0.
It is determined whether or not it is larger than (step S2301).

If the determination is NO, the tone generation circuit 108 shown in FIG. 1 instructs all the tones being sounded to be muted, and sets the variable ON NOW to indicate that there is no sound being sounded.
Is cleared (step S2306), and the sound generation process based on the zero-cross detection ends.

That is, the value of the envelope value ENV is calculated as shown in FIG.
As can be understood from the operation flowchart of FIG. 8 corresponding to step S606, the average amplitude of the input audio signal is shown. Therefore, while the average amplitude of the input audio signal is small, the sound generation processing based on the input audio signal is not performed.

The fact that the value of the note number PIT is 0 means that the note number PIT is 0 in the operation flowchart shown in FIG.
In the zero cross detection by SP116, STATUS FLAG
Does not become 4 and the substantial note number is not detected (step S621). Therefore, even in such a case, the sound generation processing based on the input sound signal is not performed.

If the determination in step S2301 in FIG. 23 is YES, it is determined whether or not the value of the variable ON NOW is 0, thereby determining whether or not there is any sound currently being generated (step S230).
2) If the determination is YES, the process proceeds to step S2305.

If the determination in step S2305 is NO and it is determined that there is a currently sounding note, the contents of the variable ON NOW indicating the note number being sounded and the contents of the variable PIT indicating the extracted note number are compared. (Step S2303).

As a result of the comparison, if they match, the tone generation process by zero-cross detection is terminated as it is not necessary to re-sound, and if they do not match, the note number to be emitted has changed. After instructing the tone generation circuit 108 to mute the currently sounding sound (step S2304), step S2305 is executed.

In step S2305, note number PIT
Is given to the tone generation circuit 108, the value of the variable PIT is substituted for the variable ON NOW, and the tone of the scale corresponding to the variable PIT is currently being produced. Is shown. After this processing, the sound generation processing based on the zero-cross detection ends.

As described above, when the pitch of the input acoustic signal is low, a musical tone is generated using scale (pitch) data obtained by zero-cross detection. Next, the sound generation processing by the filter processing in the CPU processing of FIG. 4 will be described.

As described above, in the determination at step S403 in FIG. 4, the value of the note number PIT obtained by the zero-cross detection by the DSP 116 (step S303 in FIG. 3) is above the predetermined scale K, that is, If it is determined that the pitch of the input acoustic signal is high, in step S404, the filter processing in the DSP 116 (the step of FIG.
Based on the processing result of S304), a tone generation process is executed for the tone generation circuit 108 (step S405).

FIG. 24 is an operation flowchart of the sound generation processing by the filter processing in step S404 in FIG. This operation flowchart is similar to the case of zero-cross detection,
The PU 101 is realized as an operation of executing a control program in the RAM 103.

In FIG. 24, first, the envelope signal Et (n) for N scales transferred from the DSP 116 (1 ≦ t ≦
N) is stored in the RAM 103 (step S2401).
Next, these envelope signals Et (n) (1 ≦ t ≦ N)
Among them, those whose values exceed a predetermined threshold are all extracted (step S2402).

If there is an envelope signal exceeding the threshold value (if the determination in step S2403 is YES), the one having the maximum value is extracted from them, and the scale corresponding to the envelope signal is detected (step S2403). Step S240
Four). Here, the N scales are preset in the ROM 102 or the like in advance, and the CPU 101 detects the scale by accessing the ROM 102 with the envelope signal number t having the maximum value.

Next, it is determined in the musical tone generating circuit 108 whether the extracted musical tone of the musical scale is currently being generated (step S2405). If pronunciation is in progress (Step S2
If the determination in step 405 is YES), there is no need to re-produce the sound, so that the sound generation processing by the filter processing is terminated as it is. ).

If a musical tone of another scale is not being produced (if the judgment in step S2406 is NO), the step
The tone generation circuit 108 is instructed to generate a tone based on the scale extracted in S2404 (step S2407).

If a musical tone of another musical scale is being generated, the musical tone is silenced (step S2408).
), The sound generation processing of step S2407 is performed. Step S
After the processing of 2407, the sound generation processing by the filter processing ends.

As described above, when the pitch of the input acoustic signal is high, a musical tone is generated using scale (pitch) data obtained by the filtering process.

[0164]

According to the present invention, the output of the pitch extraction system based on the zero-crossing detection and the output of the pitch extraction system based on the digital filtering are switched and selected according to the pitch frequency. It is possible to realize pitch extraction.

In particular, since the pitch extraction method using digital filtering is used only in a high pitch frequency range, the value of the filter coefficient can be reduced, and the digital filter can be easily formed by a normal DSP or the like. It becomes possible.

Further, when the pitch extracting device according to the present invention is used as a musical scale detecting device of an electronic musical instrument, the difference between musical scale frequencies in a high musical scale range becomes a large value. This eliminates the need for configuration, and as a result, the order of the bandpass filter can be reduced, and real-time processing by a DSP or the like can be easily realized.

[Brief description of the drawings]

FIG. 1 is an overall configuration diagram of an embodiment of the present invention.

FIG. 2 is an overall configuration diagram of a DSP.

FIG. 3 is an overall operation flowchart of a DSP process.

FIG. 4 is an overall operation flowchart of a CPU process.

FIG. 5 is a diagram illustrating the principle of zero-cross detection.

FIG. 6 is an operation flowchart (part 1) of zero-cross detection.

FIG. 7 is an operation flowchart (part 2) of zero-cross detection.

FIG. 8 is an operation flowchart of envelope extraction in zero-cross detection.

FIG. 9 is a basic functional block diagram of a filtering process.

FIG. 10 is a characteristic diagram of the bandpass filter Ht (z).

FIG. 11 is an improved functional block diagram of the filter processing.

FIG. 12 is a characteristic diagram of bandpass filters H1 (z) and H2t (z). is there.

FIG. 13 is a configuration diagram of a high-pass filter H1 (z).

FIG. 14 is a characteristic diagram of the high-pass filter H1 (z).

FIG. 15 is a configuration diagram of a low-pass filter H2t (z).

FIG. 16 is a diagram showing the relationship between the poles and zeros of the low-pass filter H2t (z) and the relationship between the pole vectors and the zero vectors.

FIG. 17 is an amplitude characteristic diagram of the low-pass filter H2t (z). is there.

FIG. 18 is a characteristic diagram of the low-pass filter H2t (z).
is there.

FIG. 19 is a configuration diagram of a low-pass filter H _Et (z).

FIG. 20 is a characteristic diagram of a low-pass filter H _Et (z).

FIG. 21 is a relationship diagram between | Wt (n) | and Et (n).

FIG. 22 is an operation flowchart of a filter process.

FIG. 23 is an operation flowchart of sound generation processing by zero-cross detection.

FIG. 24 is an operation flowchart of sound generation processing by filter processing.

[Explanation of symbols]

 101 CPU 102 ROM 103 RAM 104 Printer 105 Display 106 Keyboard 107 Bus 108 Musical tone generating circuit 109 Audio system 110 Speaker 111 Scale detector 112 Microphone 113 Microphone amplifier 114 Low-pass filter 115 A / D converter 116 DSP 117 ROM 201 Interface 202 Address counter 203 Operation ROM 204 Decoder 205 Multiplier 206 Flag register 207 Adder / subtractor 208 Register group

Claims (4)

(57) [Claims] 1. A zero-cross point at which a value changes from a negative or positive first polarity to another second polarity is sequentially detected from an input sound signal, and the sound signal of the sound signal is detected from an interval between the zero-cross points. First pitch detecting means for detecting pitch information; and sequentially performing digital filtering processing of different characteristics on the audio signal in a time-division manner, thereby obtaining a frequency spectrum of each of a plurality of predetermined frequencies corresponding to the processing. A second pitch detecting means for detecting a level and detecting the pitch information of the acoustic signal by determining the level of each frequency spectrum; and any one of the first pitch detecting means or the second pitch detecting means If the detected pitch information corresponds to a pitch frequency equal to or lower than a predetermined frequency threshold, the pitch information from the first pitch detecting means is output. Wherein if in response that the pitch frequency higher than a predetermined frequency threshold second
Selecting means for outputting pitch information from the pitch detecting means. 2. The apparatus according to claim 1, wherein said first pitch detecting means comprises: a zero-crossing point at which said acoustic signal crosses zero from a negative or positive first polarity to another second polarity; A first interval measuring means for measuring an interval up to a zero crossing point at which a zero crossing is made to the polarity of the first interval, after the measurement by the first interval measuring means is completed, the sound signal is converted from the first polarity to the first polarity. A second interval measuring means for sequentially measuring, as a second interval, an interval of each section from a zero-cross point at which a zero-cross to the second polarity crosses to a zero-cross point at which the second polarity crosses to the first polarity; A third interval measuring unit that sequentially measures an interval from a zero-cross point at which the measurement by the first interval measuring unit is started; and the second measuring unit starts measuring each of the second intervals. Each hour , An interval storage means for storing the interval measured by the third interval measuring means while sequentially updating the interval information as cycle information; and the first interval and the second interval measured by the first interval measuring means. Determining means for sequentially determining whether or not each of the second intervals sequentially measured by the measuring means matches within a predetermined allowable error range; and when the determining means detects the match, the time storage means The stored cycle information is output as pitch information of the sound signal, and thereafter, the pitch information is sequentially extracted from the sound signal by repeating the processing in order from the measurement processing by the first interval measuring means. The pitch extraction device according to claim 1, further comprising: a repetition control unit. 3. The method according to claim 2, wherein the second pitch detecting means performs a high-pass filtering process of a predetermined characteristic and a low-pass filtering process of adding a resonance having a peak at each of the predetermined frequencies to the audio signal. Cascade processing of division is sequentially performed, and signal processing for extracting each envelope component from each output signal corresponding to each of the predetermined frequencies obtained as a result is sequentially performed in a time division manner. Each of the predetermined frequencies obtained as a result is obtained. 3. The pitch extraction device according to claim 1, wherein a level of each of the frequency spectra is detected as each envelope component corresponding to (b). 4. An electronic musical instrument further comprising a musical tone generating means for generating musical tones of a scale corresponding to pitch information from the pitch extracting device according to claim 1.