JP3120490B2 - Voice control device and voice-controlled electronic musical instrument - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of modulating a tone signal generated from a tone generator such as an electronic keyboard with a voice signal, thereby adding an effect that the generated tone changes corresponding to the voice. More particularly, the present invention relates to a voice control device and a voice control electronic musical instrument capable of clearly generating a fricative sound of voice.

[0002]

2. Description of the Related Art With the widespread use of electronic musical instruments, there is a demand for electronic musical instruments that allow a player to more easily and effectively reflect his or her intention to play music. As one form of such an electronic musical instrument, there is an electronic musical instrument capable of modulating a tone signal by a voice signal obtained by a player uttering or the like. The present applicant has proposed an electronic musical instrument capable of adding a nuance of a human voice to the overtone component of the musical instrument by digital filter processing based on software time division processing by digital signal processing means. (Japanese Patent Application No. Hei 2-21269)
9). According to this, the spectrum of a digital voice signal based on human utterances or the like is divided into a plurality of bands by a second digital filtering process and analyzed, and the envelope signal of the spectrum for each band is subsequently executed in an envelope extraction process. Is generated. Then, a modulation process is performed in which each tone signal whose band has been limited based on the first digital filtering process is modulated by each of the envelope signals, and a modulated tone signal accumulated by the accumulation process is output. Thus, the nuance of the human voice is added to the overtone component of the instrument sound.

[0003]

However, in addition to the vocal cord sounds emitted by the vibration of the vocal cords, the human voice has a narrowed portion of the vocal tract formed by the lips, tongue and teeth, tongue and upper jaw, and the like. The white noise generated when the turbulence caused by the exhalation passes through is included in the high frequency band (called "friction sound" in academic studies). This is an important factor especially in the row and the row (see the amplitude and frequency spectra of "s" and "a" in FIGS. 16 and 17 (a) and (b)). However, there is a problem that such white noise is not included, and the nuances of the voices of the lines S and C are not properly added. SUMMARY OF THE INVENTION It is an object of the present invention to provide a voice-controlled electronic musical instrument capable of clearly adding a nuance of a human voice, particularly a "frictional" voice, to a harmonic component of a musical tone.

[0004]

According to the present invention, there is provided a white noise generating means for generating white noise, a digital signal processing means for sequentially executing the following processing at predetermined processing intervals by time-division digital signal processing, It is characterized by having. That is, first, a first digital filtering process for dividing a digital musical tone signal into musical tone signals band-limited to a plurality of different frequency bands is executed. At the same time, a second method for dividing the digital voice signal into respective voice signals band-limited to a plurality of different frequency bands.
Is performed. Here, the above-described first and second digital filtering processes are each realized as, for example, a band-pass filter process for band-limiting each input to each frequency band. , FIR-type high-pass filtering and IIR-type low-pass filtering to which resonance having a peak at the center frequency of each frequency band is added are sequentially executed in a time-division manner. Next, an envelope extraction process for extracting each envelope signal from each band-limited voice signal obtained by the second digital filtering process is executed. This process includes, for example, a low-pass filtering process having a cutoff frequency lower than the center frequency of the lowest frequency band that is band-limited in the first and second digital filtering processes.
In other words, it is realized as a low-pass filtering process that passes only the frequency components near the DC component. Further, a modulation process of modulating each band-limited tone signal obtained by the first filtering process with each envelope signal obtained by the envelope extraction process is executed. This process is realized, for example, as a process of multiplying each band-limited tone signal obtained by the first filtering process with each envelope signal obtained by the envelope extraction process. Thereafter, a third digital filtering process for limiting the white noise from the white noise generating means to a specific band is performed, and the white noise is modulated by an envelope signal of the specific band obtained by the envelope extraction process. The processing is executed. This process is realized as, for example, a process of multiplying the white noise obtained by the white noise generating means by an envelope signal of a specific band obtained by the envelope extraction process. Then, an analogization process of accumulating each modulation signal obtained by the modulation process and the noise modulation process and outputting the accumulated signal as a digital output tone signal is executed. In the above configuration, the digital tone signal can be generated from tone signal generating means, such as a digital electronic keyboard, which operates based on the performance operation by the player. Further, the digital voice signal can be generated by voice signal generating means which operates based on the operation performed by the player, for example, includes a microphone, an amplifier, an A / D converter and the like.

[0005]

The spectrum of a digital voice signal based on human utterance or the like is divided into a plurality of bands by a second digital filtering process and analyzed, and an envelope signal for each band is generated in an envelope extraction process to be executed subsequently. . Then, each tone signal whose band has been limited based on the first digital filtering process is subjected to a modulation process of being modulated by each of the envelope signals, and white noise is also reduced to a specific band, for example,
A noise modulation process that is modulated by an envelope signal of a band including a center frequency of a sound including a fricative sound such as a line is performed, and is accumulated and output by an accumulation process. Nuance can be added. In particular, in the present invention, each of the above-described processes is realized as digital filter processing based on time-division processing of software by digital signal processing means. As a result, the effect adding process of adding the nuance of the human voice to the overtone component of the instrument sound can be performed by one chip D
With SP, it can be performed easily and stably. At this time, by executing the first and second digital filter processes as a cascade connection of a high-pass filtering process and a low-pass filtering process to which resonance is added, a DSP having a relatively low processing capability is provided.
However, the real-time processing can be performed stably. Of course, if there is room in the DSP processing, the calculation processing of the bandpass filter may be configured based on the result of directly designing the transfer function.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on embodiments with reference to the drawings. FIG. 1 is an overall configuration diagram of an embodiment of the present invention. In FIG. 1, when a player performs a keyboard operation with a keyboard 33 or a switch operation such as a tone color setting or various music effect settings with a function switch 34, performance information obtained thereby is transmitted to a CPU (bus) via a bus 41. Central Processing Unit) 25. The CPU 25 executes a program stored in an R0M (Read Only Memory) 26 and executes a RAM (Random Acces
s Memory) 27 is used as a temporary work memory to process performance information. The performance information processed in this way, for example, note on / off, velocity, tone color setting data, etc., is sent to the tone generation circuit 31 via the bus 41, and the tone generation circuit 31 performs the tone generation in accordance with the above-described performance information. Is generated. As the tone generation method in the tone generation circuit 31, for example, a PCM method, a modulation method, an overtone addition method, or the like can be realized. Next, the digital tone signal x (n) generated in the tone utterance circuit 31
Is connected to a DSP (Digital
Signal Processor: Digital Signal Processor 2
8 is input.

On the other hand, when a performer or the like utters a voice toward the microphone 35, an analog voice signal obtained through the amplifier 36 is input from the low-pass filter 37 to the A / D converter 38, and converted into a digital voice signal p (n). DSP after conversion
28. Note that the analog voice signal may be input not from the microphone 35 but from the line-in terminal LINE IN.

The DSP 28 includes a filter coefficient ROM 29 storing various coefficients for a digital filtering operation to be described later, a white noise generating circuit 24 for generating white noise, and a digital tone signal x (n) and A input from a tone generating circuit 31. Work RAM 3 for storing digital voice signal p (n) input from / D converter 38 or for storing data for digital filtering operation
Using 0, an amplitude modulation process described later is executed.

The digital output tone signal z (n) obtained by the amplitude modulation processing in the DSP 28 is sent to the D / A converter 32 via a dedicated bus 43, where it is converted into an analog output tone signal. After being amplified by the amplifier 39, the sound is emitted from the speaker 40.

FIG. 2 is a block diagram of the DSP 28 shown in FIG. First, the interface 281 includes a bus 22 connected to the white noise generation circuit 24, a bus 23 connected to the A / D converter 38, and a bus 4 connected to the CPU 25.
1. Bus 42 connected to the tone generator 31 and D / A
A bus 43 connected to the converter 32 is accommodated, and each bus is connected to an internal circuit.

The operation ROM 282 stores the DSP 2
8 is a ROM in which a microprogram for controlling the entire operation is stored, and a corresponding program instruction is read out based on a designated address from the address counter 283. The CPU 25 of FIG.
By setting a value in the address counter 283, an instruction is given as to what program instruction is to be read from and the modulation process described later is to be executed. Operation ROM2
The output of 82 is also provided to decoder 284, and D
Various control signals are output to each circuit in the SP 28 to perform a desired operation.

A filter coefficient ROM 29 and a work RAM 30 shown in FIG. 1 are connected to the internal bus of the DSP 28, and filter coefficient data and digital tone signals x are appropriately adjusted according to program instructions of an operation ROM 282.
(n), a digital voice signal p (n) or the like is supplied to the DSP 28 or input / output to / from the work RAM 30.

The DSP 28 further has a multiplier 285 and an adder / subtractor 286 for arithmetic processing, and is connected to the internal bus in the form of two inputs and one output, respectively. The register group 287 includes a plurality of registers for storing data in the middle of a calculation, and is connected to each input / output terminal of the multiplier 285 or the adder / subtractor 286 via an internal bus.

The DSP 28 sends a flag signal indicating the judge result to the address counter 283 via the flag register 288 in order to perform a judge process according to the operation result (comparison result or the like) from the adder / subtracter 286. The address value of the address counter 283 is appropriately changed according to the output of the flag register 288, and the program command whose address has been jumped is read from the operation ROM 282.

FIG. 3 is a block diagram showing functions realized by the DSP 28. As shown in FIG. A band-by-band converter 441 realized by time-division processing of software on the DSP 28;
, 44 _i = _j ,..., 44 _N operate at each sampling period, and at the end of each sampling, an output from each conversion unit is accumulated by a software processing on the DSP 28. Digital output tone signal z (n)
Is output to the D / A converter 32 in FIG.

Each band converting section 44 _t (1 ≦ t ≦ N) further includes band pass filter sections (BPF sections) 45 and 46, an envelope extracting section 47 and a multiplying section 48. BPF
The units 45 and 46 are realized by a combination of a high-pass filter by software processing common to each band and a low-pass filter by software processing for each band, as described later.

The envelope extracting section 47 is realized by a low-pass filter by software processing having a low cutoff frequency, as described later. The multiplying unit 48 is realized by a product-sum operation process together with the accumulating unit 49, as described later.

The specific band by converting portion 44 _m in the respective bands by conversion unit _{44 1 ~44 N (1 ≦ m} ≦ N) , said band-pass filter portions 45 and 46, the envelope extraction unit 47 and the multiplication section 48 And a band-pass filter unit 80 and a multiplication unit 81.

Next, the detailed principle configuration of the BPF units 45, 46, 80 and the envelope extracting unit 47 in FIG. 3 will be described. The digital tone signal x (n) and the digital voice signal p at each sampling timing n input from the tone generator 31 and the A / D converter 38 in FIG.
(n) shows N BPs by the time division processing of the DSP 28.
Filtering processing is executed in the F units 45, 46, and 48. BPF units 45 and 4 in each band-specific conversion unit 44 _t
6 Then, BPF portion of the specific band by converting portion 44 _m 80,
Have the same transfer function, and the transfer function is H
_{Let t} (z). In this embodiment, the BPF unit is realized by cascade connection of a high-pass filter unit common to each band and a low-pass filter unit for each band. Assuming that the transfer function of the high-pass filter unit is H ₁ (z) and the transfer function of the low-pass filter unit is H _2t (z), the transfer function H _t (z) becomes H ₁ (z) and H _{2t as} shown in FIG. The characteristic is represented by the product of (z).

In the case of the BPF unit 45 shown in FIG.
The digital tone signal x (n) is filtered by the high-pass filter of the transfer function H ₁ (z) and then transferred to the transfer function H (z).
A digital tone signal Y _i (n) filtered by a low-pass filter section of _2t (z) and band-limited (where i =
t).

In the case of the BPF section 46, the digital tone signal p (n) is filtered by the high-pass filter section of the transfer function H ₁ (z) and then filtered by the low-pass filter section of the transfer function H _2t (z). , And is output as a digital voice signal Q _j (n) (i = t) whose band is limited. Further, the digital voice signal Q _j (n) is subjected to processing in the envelope extraction unit 47 in FIG. 3, and this part has a cutoff frequency having a transfer function H _Et (z) as shown in FIG. This is realized by a low-pass filter unit. With such a low-pass filter section, an envelope signal R _j (n) is obtained from the digital voice signal Q _j (n).

In the case of the BPF unit 46, the white noise signal W (n) is filtered by the high-pass filtering unit of the transfer function H ₁ (z), and then the transfer function H ₂ (z).
, And output as a band-limited digital white noise signal U _m (n).

The characteristics of the high-pass filter section of the transfer function H ₁ (z) and the low-pass filter sections of the transfer functions H _2t (z) and H _Et (z) will be described in detail below.

FIG. 5 is a block diagram showing the hardware of the high-pass filter H ₁ (z) of FIG. This is a second-order FIR digital filter, and the transfer function is H ₁ (z) = (1/4) (1-2Z ⁻¹ + Z ⁻² ).

In FIG. 5, reference numerals 50 and 51 denote delay elements, 5
2, 53 and 54 are multipliers, and 55 and 56 are adders. In the DSP 28 (FIG. 1 or FIG. 2), a filter operation equivalent to the high-pass filter having the configuration of FIG.
In the case of the BPF unit 45 (FIG. 3), S (n) = (1/4) (p (n) -2x (n-1) + x (n-2)) (1) In the case of the BPF unit 46 (FIG. 3), S (n) = (1/4) (p (n) -2p (n-1) + p (n-2)) (2) This is realized by discrete arithmetic processing. In this case,
Since the filter coefficient is a multiple of 2, the multiplication of the coefficient and the signal can be realized by simple bit shift processing.

The frequency characteristics of this high-pass filter are as follows:

(Equation 1) It becomes minimum at Ω = 0 (0Hz), having a _{Ω = π (f s / 2Hz} ) becomes maximum in characteristics. Here, f _s is the sampling frequency (common) of the digital tone signal x (n) and the digital voice signal p (n). FIG. 6 shows the characteristics.

FIG. 7 shows the low-pass filter H _2t (z) of FIG.
FIG. 1 is a configuration diagram showing a hardware image. This is a second-order IIR digital filter whose transfer function is It is. As described later, each band-specific conversion unit 44 _t in FIG.
Θ and CY change depending on the subscript t, and r is a parameter indicating the intensity of resonance (degree of peak).

In FIG. 7, 57 and 58 are delay elements, 5
9, 60 and 61 indicate multipliers, and 62 and 63 indicate adders. In the DSP 28 (FIG. 1 or FIG. 2), a filter operation equivalent to the low-pass filter having the configuration of FIG. 7 is performed as follows: W _t (n) = CY · S (n) + 2rcos θW _t (n−1) −r ² W _t (n-2)... (3) It is realized by the discrete operation processing as follows.

[0029]

(Equation 2)

Here, the pole of the transfer function is

[Outside 1] And there is a double zero at z = 0. FIG. 8 shows the arrangement of the poles and zeros and the relationship between the pole vector and the zero point vector when 0 <θ <π / 2. As can be understood from the figure, as Ω moves along the unit circle from Ω = 0 to Ω = π, the length of the vector v2 first decreases and then increases. The length of the smallest vector v ₂ is the near pole (rej). Here, the magnitude of the frequency response at the frequency Ω is
The ratio of the zero point vector v ₁ pole vector v ₂ length, frequency response phase, it becomes a value obtained by subtracting pole angle of the vector v ₂ from the real axis and the zero point angle of the vector v ₁ is known FIG. 9 shows only the amplitude characteristics. That is, the size (amplitude characteristics) of the frequency response, poles proportional to the magnitude inverse of the vector v _2, it can be seen that the maximum at near theta Omega. Then, the sharpness of this peak is determined according to the magnitude of r, and as r approaches 1, a filter having a steep peak (resonance characteristic) can be realized.

As is apparent from the above description, if determining the value of theta for each respective band converting unit 44 _t of FIG. 3 (θ = 2πf
_t / f _s), as shown in FIG. 10, the center frequency of each band
A low-pass digital filter with resonance having a peak at f _t can be realized.

Here, r is sized so as not to affect the adjacent band, CY is sized empirically or mathematically to obtain an output W _t (n) of the same level in each band. It is possible to ask for. For example, the center frequency f of each band
Assuming that the ratio of the magnitude of the frequency response between _t and the center frequency f _t + Δf (that is, f _{t + 1} ) of the adjacent band Δf is m: 1,

(Equation 3) Solve the quartic equation for r that, to choose those 0 <r <1, each coefficient -2Rcosshita, it is possible to obtain the r ^2. As a result of numerical calculation, for example, sampling frequency f _s = 5k
Hz, f = 440Hz, and m = 4, -2rcosθ = 1.9773, r ² = 0.
9851, CY = 36.7. The same applies to other bands.

A high-pass filter having the transfer function H ₁ (z) as described above and a low-pass filter having the transfer function H _2t (z) are cascaded as shown in FIG. By the overall transfer function, expressed as the product of each transfer function, as shown in FIG.
A pseudo bandpass filter having center frequencies f _{1 to} f _N and a frequency difference Δf between adjacent bands can be realized for each band of = 1 to t = N.

Next, FIG. 12 is a configuration diagram showing a hardware image of the low-pass filter H _Et (z) of FIG. 4 corresponding to the envelope extraction unit 47 of FIG. This is a second-order IIR digital filter having the same form as the low-pass filter H _2t (z) described above, and the transfer function is It is. This is obtained by setting r = 0.9 and θ = 0 in the transfer function of the low-pass filter H _2t (z).

In FIG. 12, reference numeral 64 denotes an absolute value circuit for converting the output Q _j (n) of the BPF section 46 of FIG. 3, which is the output W _t (n) of the low-pass filter H _2t (z) of FIG. 4, into an absolute value. And its output | Q _j (n) | is digitally filtered. Reference numerals 65 and 66 denote delay elements, reference numerals 67, 68 and 69 denote multipliers, and reference numerals 70 and 71 denote adders. In the DSP 28 (FIG. 1 or FIG. 2), a filter operation equivalent to the low-pass filter having the configuration of FIG. 12 is performed as follows: R _j (n) = CE | Q _j (n) | + 1.8R _j (n−1) − 0.81R _j (n-2)... (4) It is realized by the discrete operation processing as follows.

As is clear from the above description, the frequency characteristic of this low-pass filter is a resonance-equipped low-pass filter having a peak at θ = 0, and has the characteristic (amplitude characteristic) shown in FIG. In this case, the cutoff frequency is set to a lower frequency than the cutoff frequency of the low-pass filter H _2t (z) in the lowest band. Here, the coefficient CE is a coefficient for making the level of each band uniform, and can be empirically obtained as appropriate.

FIG. 14 is a diagram schematically showing the envelope signal R _j (n) obtained according to the above principle in comparison with the input | Q _j (n) |. By the operation corresponding to the absolute value circuit 64 in FIG. 12, all the negative peak values (broken lines in FIG. 14) are converted into positive peak values, and then low-pass filtering is performed. The operation of obtaining the DC component of the digital voice signal | Q _j (n) | is performed by the envelope extracting unit 47 of FIG.

The operation when the filter function shown in FIGS. 3 to 14 is executed as software processing by the DSP 28 shown in FIG. 1 or 2 will be described below.

FIG. 15 is a flowchart showing the operation of the DSP in accordance with the microprogram stored in the operation ROM 282 (see FIG. 2), ie, the band-pass filter processing and the low-pass filter processing for envelope extraction, that is, the DSP vocoder processing. .
Thus, for each sampling period common to the digital tone signal x (n), the digital voice signal p (n), and the white noise signal W (n), the band-by-band conversion unit 44 _t (t
= 1 to N), the processing corresponding to the BPF units 45, 46, 80, the envelope extracting unit 47, the multiplying units 48, 81, and the processing corresponding to the accumulating unit 49 are executed as time-division processing. A digital output tone signal z (n) is obtained for each sampling period, and the D / A converter 32 shown in FIG.
Is output to

In FIG. 15, first, the contents of the work RAM 30 shown in FIG. 1 or 2 are initialized (step S1).

Next, wait for the A / D conversion completion signal generated at a period corresponding to the sampling frequency f _s from the A / D converter 38 of FIG. 1 (step S2), the each sampling timing, A / D The converted digital voice signal p (n) is taken in from the interface 281 (see FIG. 2), and is sequentially stored in the work RAM 30. At this time, the digital tone signal x (n) input from the tone generating circuit 31 (see FIG. 2) is fetched from the interface 281 and sequentially stored in the work RAM 30. The white noise signal W (n) input from the white noise generation circuit 24
From the interface 281 and the work RAM 30
Are stored sequentially. The work RAM 30 can store a current sample and two past samples.

Next, the process of step S4~S9 in Figure 15, the BPF 46 of the respective band converting unit 44 ₁ ~ 44 _N of FIG. 3
And the processing of the envelope extraction unit 47. That is, first, from the work RAM 30, the register group 28 of FIG.
The current digital voice signal and the digital voice signals of the past two samples are read out to the registers p (n), p (n-1), and p (n-2) in 7, and are stored in the BPF unit 46 in FIG. A high-pass filtering process represented by the transfer function H ₁ (z) in FIG. 4 that is a part of the corresponding process is executed (Step S4 in FIG. 15). This processing is an arithmetic processing represented by the above-described equation (2), and is executed using the multiplier 285 and the adder / subtractor 286 in FIG. At this time, each filter coefficient used for the calculation of the expression (2) is read from the filter coefficient ROM 29 (see FIG. 1 or 2) and used for the calculation. The resulting output is stored in register S (n) in register group 287. Since this high-pass filtering process is a process common to each band, it may be performed only once.

Next, a low-pass filtering process represented by the transfer function H _2t (z) = H _2j (z) in FIG. 4, which is the rest of the process corresponding to the BPF unit 46 in FIG. Transfer processing H of FIG.
Low-pass filtering processing represented by _Et (z) = H _Ej (z) is continuously performed. These processes correspond to the respective band converting unit 44 ₁ ~ 44 _N of FIG. 3 are repeated as time division processing of N bands minute. Therefore, the register group 28 shown in FIG.
7 is provided with a register j for repetition control for performing N-band time-division processing, is initialized to a value of 1 in step S5, and the low-pass filtering processing for one band in steps S6 and S7 ends. Step S8 for each
It is determined whether or not the contents of the register j have reached N.
If not, the content of the register j is incremented in step S9, and step S6 is repeated. This processing is executed by the adder / subtractor 286 and the flag register 288 shown in FIG.
Causes the program instructions corresponding to steps S6 and S7 to be repeatedly read from the operation ROM 282.

First, the transfer function H of FIG. 4 is applied to the contents of the register S (n), which is the output of the above-described high-pass filter processing.
A low-pass filtering process represented by _2t (z) = _H2j (z) is executed (step S6 in FIG. 15). This process is an arithmetic processing expressed as W _t = Q _j in the above (3) (Fig. 4
2) and the adder / subtractor 286 of FIG. Each filter coefficient used in the calculation processing of the equation (3) is read from the filter coefficient ROM 29 and used for the calculation. In the register group 287, there are provided registers Q _j (n-1) and Q _j (n-2) for storing their own filter outputs for the past two samples. Used for The resulting output is stored in register Q _j (n) in register group 287. Note that each register Q _j (n), Q _j (n-1) and Q _j (n-2)
N bands are provided by changing the suffix j.

Next, the contents of an output register Q _j of the low-pass filter process described above (n), low-pass filtering process represented by transfer function H _Et in FIG _{4 (z) = H Ej (} z) The process is executed (Step S7 in FIG. 15). This processing is an arithmetic processing represented by the above-described equation (4), and is executed using the multiplier 285 and the adder / subtractor 286 in FIG.
Also in this case, each filter coefficient used for the calculation processing of the equation (4) is read from the filter coefficient ROM 29. Also, in the register group 287, registers R _j (n−1) for storing own filter outputs for the past two samples and R
_j (n-2) are provided, and their contents are also used in the above operation. The resulting output is stored in register group 287
_Is stored in the register R _j (n). Note that each register R
_j (n), R _j (n-1) and R _j (n-2) are obtained by changing the subscript j to N
It is provided for the band.

The above, by the processing of step Step S5~ step S9 in FIG. 15, the processing corresponding to the N band portion of the band-specific conversion unit 44 ₁ ~ 44 _N BPF unit 46 and envelope extractor 47 of FIG. 3 is executed You.

Subsequently, steps S10 to S13 in FIG.
Process, each BPF of respective band conversion unit four hundred forty-one to forty-four _N of FIG. 3
This corresponds to the processing of the unit 45.

That is, first, the registers x (n), x (n-1), x (n-n) in the register group 287 of FIG.
In 2), digital tone signals of the present and the past two samples are read out, and a high-pass filtering process indicated by a transfer function H ₁ (z) in FIG. 4 which is a part of the process corresponding to the BPF unit 45 in FIG. Is executed (step S1 in FIG. 15).
0). This processing is an arithmetic processing represented by the above-described equation (1), and is executed using the multiplier 285 and the adder / subtractor 286 in FIG. At this time, each filter coefficient used for the calculation of the expression (1) is read from the filter coefficient ROM 29. The resulting output is stored in register S (n) in register group 287. Since this high-pass filtering process is a process common to each band, it may be performed only once.

Next, a low-pass filtering process represented by a transfer function H _2t (z) = H _2i (z) in FIG. 4, which is the rest of the process corresponding to the BPF unit 45 in FIG. 3, is executed. This process corresponds to the respective band converting unit 44 ₁ ~ 44 _N of FIG. 3 are repeated as time division processing of N bands minute. For this purpose, a register i for repetition control for performing time division processing of N bands is provided in the register group 287 of FIG. 2, and is initialized to a value of 1 in step S11. Thereafter, each time the low-pass filtering process for one band in step S12 is completed, it is determined in step S13 whether or not the content of the register i has reached N.
At 14, the contents of the register i are incremented, and step S12 is repeated. Also in this case, each circuit in FIG. 2 operates similarly to the case of the register j described above.

The processing in step S12 is almost the same as the processing in step S6 described above. That is, the contents of the register is the output of high-pass filtering S (n), the arithmetic processing is performed expressed as W _t = Y _i in the above (3) (see FIG. 4). At this time, registers Y _i (n−1) and Y _i (n−2) for storing their own filter outputs for the past two samples are provided in the register group 287, and their contents are also described above. Used for arithmetic. The resulting output is stored in register Y _i (n) in register group 287. Note that each register Y _i (n), Y _i (n-1) and Y _i (n-2)
N subbands are provided by changing the subscript i.

As described above, by the processing in steps S10 to S13 in FIG. 15, the band-by-band conversion units 44 ₁ to 44 for N bands in FIG.
4 process corresponding to the BPF section 45 of _N is performed.

As described above, in steps S4 to S13, FIG.
, The contents of the registers R _j (n) and Y _i (n) corresponding to the outputs of the envelope extractor 47 and the BPF 45 are determined. Using these contents, the band-specific converters 44 for the N bands in FIG.
_The following processing corresponding to the processing of the accumulating unit 49 in FIG. 3 is executed similarly to the multipliers 48 of _{1 to} 44 _N. That is, FIG.
In step S15 of 5, the multiplication of R _j (n) × Y _i (n) is performed while changing the contents of the register i = j from 1 to N. This is done using the multiplier 285 of FIG. Then, the respective multiplication results are accumulated. This is performed using the adder / subtractor 286 of FIG.

The accumulation result obtained as described above is stored in the register z (n) in the register group 287 in FIG.
Subsequently, the specific band conversion unit 44 _{m of} the white noise signal W (n)
The processing corresponding to the BPF unit 80 and the multiplication unit 81 is performed in steps S16 to S18.

That is, first, the work RAM 30 is
Registers W (n), W (n-1), W (n-2) in the register group 287 of FIG.
Then, the white noise signals of the current and past two samples are read out, and the high-pass filtering process indicated by the transfer function H ₁ (z) in FIG. 4 which is a part of the process corresponding to the BPF unit 80 in FIG. 3 is executed. Is performed (step S16). This processing is an arithmetic processing represented by the above-described equation (1), and is executed using the multiplier 285 and the adder / subtractor 286 in FIG.
At this time, each filter coefficient used for the calculation of the expression (1) is read from the filter coefficient ROM 29. The resulting output is stored in register S (n) in register group 287.

Next, a low-pass filter process represented by the transfer function H _2t = H _2m (z) in FIG. 4, which is the rest of the process corresponding to the BPF unit 45 in FIG. 3, is executed. This processing is performed only for a specific band unlike the tone signal and the voice signal (step S17).

Then, the white noise signal U _m (n), which has been low-pass filtered in step S17, is then multiplied by the envelope signal R _m (n) extracted in the same specific band m (step S18). Signal V _m (n)
Is added to the register z (n) in the register group 287 and stored again in the register z (n) (step S19). In the subsequent step S20 of FIG. 15, at the sampling interval, the D / A of FIG.
Output to the converter 32.

According to the embodiment described above, the BPF unit 4 shown in FIG. 3 for dividing a musical tone signal and a human voice signal into a plurality of bands.
5, processing of 46, the processing of the envelope extraction unit 47 of FIG. 3 for extracting the envelope from the band-limited voice signal, the processing of the multiplication unit 48 of FIG. 3 for applying amplitude modulation to the tone signal band-limited by the envelope signal, The BPF unit 80 limits the noise signal to a specific band, the multiplication unit 81 performs amplitude modulation of the band-limited white noise signal with an envelope signal, accumulates the modulation output of each band, and outputs an output tone. The processing of the accumulator 49 in FIG. 3 for obtaining the signal is realized as a digital voice signal by software time-division processing. Thus, the effect adding process of adding the nuance of the human voice to the overtone component of the musical tone signal and also clearly adding the nuance of the human voice including the fricative sound is simply and stably performed by the one-chip DSP. be able to.

If there is room in the DSP processing, the operation of the band-pass filter is not performed as the operation of the combination of the high-pass filter and the low-pass filter as in the above-described embodiment, but the transfer function of the band-pass filter is used. May be realized by arithmetic processing configured based on the result of direct design.

Further, the processing may be performed by DSP without specially providing the white noise generating circuit 24 as in this embodiment.

[0060]

According to the present invention, the nuances of the human voice including fricative sounds can be clearly added to the overtone components of the musical instrument sound by digital filter processing based on software time division processing by the digital signal processing means. A possible voice control device and voice control electronic musical instrument can be easily realized simply by adding one or several chips of DSP to a conventional digital electronic musical instrument. This makes it possible to provide a low-cost voice-controlled electronic musical instrument with a good S / N ratio and little variation in performance over time or mass production.

In particular, by executing the first and second digital filter processes as a cascade connection of a high-pass filtering process and a low-pass filtering process to which resonance is added, even a DSP having a relatively low processing capability can perform real-time processing. Processing can be performed stably.

[Brief description of the drawings]

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

FIG. 2 is a configuration diagram of the DSP of FIG. 1;

FIG. 3 is a functional block diagram of a DSP.

FIG. 4 is a filter configuration diagram of a BPF unit and an envelope extraction unit.

FIG. 5 is a configuration diagram of a high-pass filter H ₁ (z).

FIG. 6 is a characteristic diagram of a high-pass filter H ₁ (z).

FIG. 7 is a configuration diagram of a low-pass filter H _2t (z).

FIG. 8 is a diagram _showing the relationship between a pole and a zero point and the pole vector and a zero vector of the low-pass filter H _2t (z).

FIG. 9 is an amplitude characteristic diagram of a low-pass filter H _2t (z).

FIG. 10 is a characteristic diagram of a low-pass filter H _2t (z).

FIG. 11 is a characteristic diagram of the band-pass filters H ₁ (z) and H _2t (z).

FIG. 12 is a configuration diagram of a low-pass filter H _Ej (z).

FIG. 13 is a characteristic diagram of a low-pass filter H _Ej (z).

FIG. 14 is a relationship diagram between | Q _j (n) | and R _j (n).

FIG. 15 is an operation flowchart of a DSP vocoder process.

FIG. 16 is an amplitude waveform of a single sound “sa”.

FIG. 17 is a frequency spectrum waveform of a single sound “sa”.

[Explanation of symbols]

 24 White noise generating circuit 28 DSP 29 Filter coefficient ROM 30 Work RAM 281 Interface 282 Operation ROM 283 Address counter 284 Decoder 285 Multiplier 286 Adder / Subtractor 287 Register group 288 Flag register

Claims (8)

(57) [Claims] 1. A white noise generating means for generating white noise, a first digital filtering process for dividing a digital tone signal into tone signals band-limited in a plurality of different frequency bands, and a digital voice signal A second digital filtering process of dividing into voice signals band-limited within different frequency bands, an envelope extraction process of extracting each envelope signal from each band-limited voice signal obtained by the second digital filtering process, A modulation process of modulating each band-limited tone signal obtained by the first filtering process with each envelope signal obtained by the envelope extraction process; and converting white noise generated by the white noise generating means into a specific frequency band. White noise band-limited A third digital filtering process for performing signal processing on the signal, a noise modulation process for modulating a white noise signal obtained by the third digital filtering process with an envelope signal in a specific band obtained by the envelope extraction process, and the modulation Processing and an accumulation process for accumulating each modulated signal obtained in the noise modulation process and outputting it as a digital output tone signal; and a digital signal process for sequentially executing at predetermined processing intervals by time-division digital signal processing. Digital signal processing means for sequentially executing processing at predetermined processing intervals by time-division digital signal processing. 2. The digital signal processing means according to claim 1, wherein said first and second digital filtering processes are respectively performed as band-pass filter processes for limiting an input to each of said frequency bands. The voice control device according to claim 1. 3. The digital signal processing means according to claim 1, wherein said first and second digital filtering processes include a high-pass filtering process, and a low-pass filtering process to which resonance having a peak at a center frequency of each of said frequency bands is added. 3. The voice control device according to claim 2, wherein the processes are sequentially performed in a time-division manner. 4. The digital signal processing means according to claim 1, wherein said first and second digital filtering processes are respectively performed by an FIR type high-pass filtering process and an IIR to which a resonance having a peak at a center frequency of each frequency band is added. 3. The voice control device according to claim 2, wherein the low-pass filtering process of the type is sequentially performed in a time-division manner. 5. The digital signal processing means according to claim 1, wherein said envelope extraction processing is a low-pass filtering processing having a cutoff frequency lower than a center frequency of a lowest frequency band which is band-limited in said first and second digital filtering processing. The voice control device according to claim 1, wherein the voice control device is executed as: 6. The voice control apparatus according to claim 1, wherein said digital signal processing means executes said envelope extraction processing as a low-pass filtering processing for passing only a frequency component near a DC component. 7. The digital signal processing means performs the modulation process as a process of multiplying each band-limited tone signal obtained by the first filtering process by each envelope signal obtained by the envelope extraction process. The voice-controlled electronic musical instrument according to claim 1, wherein the electronic musical instrument is executed. 8. A tone signal generating means for generating a digital tone signal based on a performance operation by a player, a voice signal generating means for generating a digital voice signal based on a vocal operation by the player, and generating white noise. White noise generating means, digital signal processing means for performing digital signal processing to output a digital output tone signal, and converting the digital output tone signal output from the digital signal processing means into an analog output tone signal to emit sound Sound emitting means, first digital filtering processing for dividing a digital tone signal generated by the tone signal generating means into tone signals which are band-limited to a plurality of different frequency bands, and a digital signal generated by the voice signal generating means. Divide the voice signal into each band-limited voice signal within several different frequency bands 2 digital filtering processing, envelope extraction processing for extracting each envelope signal from each band-limited voice signal obtained by the second digital filtering processing, and each band-limited tone signal obtained by the first filtering processing Is modulated by each envelope signal obtained by the envelope extraction processing, a third digital filtering processing for band-limiting a white noise signal generated by the white noise generating means to a specific band, the third digital A noise modulation process of modulating white noise obtained by the filtering process with an envelope signal of a specific band obtained by the envelope extraction process, accumulating each modulation signal obtained by the modulation process, and outputting a digital output tone signal Cumulative output as Voice control electronic musical instrument characterized by having processes, a digital signal processing means for sequentially executing a predetermined processing interval by a digital signal processing at the time of dividing each process, the.