JPH06102882A - Method for composing musical sound signal - Google Patents

Abstract

PURPOSE:To obtain the method for composing the musical sound signal whose data are compressed so as to reduce the capacity of a memory that the sound source of an electronic musical instrument is equipped with. CONSTITUTION:The memory is stored with initial setting data on amplitude component between the amplitude components and phase components by discrete frequencies obtained on the basis of a specific equal number of data by discrete frequencies found by discrete Fourier transformation by using the same window function for signals of respective sequential Fourier transformation frames of specific time length cut from a musical sound signal, initial setting data on the phase components, and data on a value obtained by rounding the value thetaa of an arithmetic term, whose phase thetam of specific discrete frequency in an (m)th Fourier transformation frame is represented as thetam=thetam-1+(theta-1-thetam-2)=thetam-1+thetaa when the phase of the specific discrete frequency in an (m-1)th Fourier transformation frame cut from the musical sound signal so as to have specific time length is thetam-1 and the phase of the specific discrete frequency in an (m-2)th Fourier transformation frame is thetam-2 as to the phase components by the discrete frequencies obtained by the Fourier transformation frames, to 2pi/n (n: integer).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of synthesizing tone signals.

[0002]

2. Description of the Related Art In recent years, data obtained by digitizing a musical tone signal generated from an actual musical instrument or the like is stored in a memory, and the memory storing the data is used to generate a musical tone. Electronic musical instruments and other devices having a configuration as provided in a sound source have come into wide use. In order to perform a musical performance close to that of an actual musical instrument by the above-described electronic musical instrument or the like, it is necessary that the data correspond to a large number of timbres. Since a large-capacity memory is required when configuring a device such as an electronic musical instrument having data corresponding to a timbre, the form of the device becomes large and the cost becomes high.

The above-mentioned problem can be solved by, for example, highly efficient encoding data for each timbre, compressing the amount of data, and storing it in a memory. According to Japanese Patent Application No. 4-30076 “Acoustic Signal Phase Prediction Method”, the first and the second in each successive Fourier transform frame cut out from the acoustic signal so as to have a predetermined constant time length. The same predetermined number of data for each discrete frequency obtained as a result of the Fourier transform for each of the first and second Fourier transform frames described above by discretely Fourier transforming the Fourier transform frame using the same window function. Thus, the phase information for each discrete frequency is obtained for each of the first and second Fourier transform frames described above, and the phase information is obtained for each of the first and second Fourier transform frames described above. The corresponding change mode of the phase information for each discrete frequency is obtained, and the above-mentioned first time position is assumed, assuming that the change mode of the phase information for each discrete frequency is constant on the time axis. And the second time position, an individual multiple of a predetermined number of discrete frequencies in the Fourier transform frame at the third time position existing at a time position that is an integer multiple of the time difference between the third time position and the third time position is determined. A method for predicting the phase information of the Fourier transform frame of the position has been proposed, and by applying the above-described method for predicting the phase information, high-efficiency encoding of the acoustic signal is performed, thereby transmitting (recording) the acoustic signal, The configuration of the sound source of was made easy.

FIG. 4 is a block diagram of an encoder which can be used also as a means for compressing data for a tone sound source, which is one of the application examples of the previously proposed method of predicting phase information by the applicant company, FIG. 5 is a block diagram of the decoder. Further, FIG. 6 is a diagram which is also used for explaining the configuration principle and the operation principle of the previously proposed method of predicting phase information, and the AU described in the uppermost part in FIG.
The waveform labeled DIO is an acoustic signal that is targeted for high efficiency encoding. The acoustic signal in FIG. 6 is Fra
As indicated by 0, 1, 2, ... Shown in a part of the display of me window, each is cut out so as to have a constant time length, and is made a sequential Fourier transform frame. A state in which each Fourier transform frame described above is multiplied by a window function so as to overlap each other so as to have a period having, for example, 1024 sampling points from an acoustic signal and to be gradually connected to each other. Of the Fourier transform frame for each frame described above is subjected to orthogonal transform by, for example, discrete Fourier transform (DFT) or fast Fourier transform (FFT). Be done. In the following description, the above orthogonal transformation is described as being performed by a fast Fourier transform (FFT).

When the FFT operation is performed for each Fourier transform frame for each frame, the number of data (sample number) in the Fourier transform frame for each frame is N, and the sampling frequency is fs. , F
= Fs / N, the FFT calculation result is obtained for each discrete frequency (total of N frequencies) for each frequency interval of f, and the FFT calculation result is a real number for each discrete frequency. Part (Real) amplitude and imaginary part (Ima
g) amplitude. By using the real part (Real) amplitude and the imaginary part (Imag) amplitude for each discrete frequency described above, the following Formula 1 and Formula 2 are used to perform polar coordinate conversion for each discrete frequency described above. Composite amplitude term (Am
p) and the phase term (Phase). By the way, FF
Of the N real part (Real) amplitudes and the N imaginary part (Imag) amplitudes obtained for each discrete frequency as a result of the T calculation, the N real part (Real) amplitudes have the same value. Two of them appear, and there are N imaginary parts (I
mag) Since two amplitudes each having the same absolute value appear, the number of valid combined amplitude terms (Amp) is 1/2 of the total number, and the valid phase term (Phase) terms are also included. Since the number is 1/2 of the total number, the number of FFT output data is F
It can be equal to the number of real input data of FT.

[0006]

[Equation 1]

[Equation 2]

Now, for successive Fourier transform frames (frames) continuous on the time axis, the real part (Real) amplitude and the imaginary part (Ima) of each discrete frequency.
g) Amplitude and by equation 1 and equation 2, the combined amplitude term (Am
When p) and the phase term (Phase) are obtained, it is natural to predict that two adjacent frames on the time axis will have the same amplitude if the most common sense is taken. Yes, and actually, for example, the sampling frequency is 44.1
Assuming that the sample length of 1024 points is set as the frame length (frame length is about 1/40 seconds) at KHz, the spectrum obtained by the FFT operation result obtained when the FFT operation is performed on the signal of the piano is also 1 Comparing 512 amplitudes in one frame with 512 amplitudes in the next frame of the frame, or comparing 512 amplitudes in the next frame of the next frame, in different frames It is also confirmed that the amount of change between spectra is extremely small.

On the other hand, it may be difficult to predict the phase in successive frames that are continuous on the time axis, because the repetition time of successive frames and the frequency of the signal are irrelevant, and the phase of the signal It is clear from the viewpoint that the frame starts and ends regardless of the above, and it has been conventionally considered difficult to perform signal prediction by orthogonal transform. Then, assuming that the sampling frequency is 44.1 KHz and the number of samples is 1024 points as the frame length,
Looking at the distribution of 512 phases in one frame based on the FFT operation result obtained by actually performing FFT operation on the piano sound signal, the distribution of the phases is random. 5 in the frame
It was found that the distribution of 12 phases cannot be predicted.

[0009] Mr. Takahashi, the inventor of the above-mentioned already proposed “method of predicting the phase of an acoustic signal” by the applicant company, says that even if the phase information for one frame in a sequential frame on the time axis is used, It is not possible to predict the phase information of each frame, but for the two frames, the phase information between the phase information of the same predetermined number of discrete frequencies obtained as a result of the Fourier transform of each Fourier transform frame is calculated. Paying attention to the fact that if the change mode is constant on the time axis, it is possible to predict the phase information of other frames by using that relationship. The change mode of the phase information between the phase information of the same predetermined number of discrete frequencies obtained as a result of the Fourier transform for each transform frame is uniform on the time axis. It is based on the hypothesis (hereinafter referred to as Takahashi's hypothesis) that a sine wave signal of a single frequency, a composite signal of sine wave signals of multiple frequencies, a signal of a musical instrument (piano) sound, After conducting experiments using various signals such as, the phase of the frame predicted according to the above hypothesis and the phase of the actual frame are correct, and the correct prediction results are obtained to the extent that they are recognized to be practically consistent. The fact that Takahashi's hypothesis holds in practice is supported by various experimental results.

According to Takahashi's hypothesis, data θi (1) of N / 2 phase terms obtained for each discrete frequency in frame 1 shown in FIG. Data θi (2) of N / 2 phase terms obtained for each discrete frequency and data θi (3) of N / 2 phase terms obtained for each discrete frequency in frame 3, for example. ) And, for example, N / 2 phase term data θi (4) obtained for each discrete frequency in frame 4, the phase term data in frame 1 is the phase term data in the same frequency value. Data is, for example, θ1
If the data of the phase term in frame 2 is, for example, θ2, the data of the phase term in frame 3 is, for example, θ3, and the data of the phase term in frame 4 is, for example, θ4. ,
Since the amount of change in phase between frames is almost the same as θ2-θ1 ≈ θ3-θ2 ≈ θ4-θ3 ≈ Δθa, if this hypothesis holds, then for two frames Knowing the phase difference between the data of the phase terms of all the frequency values corresponding to each other in the N / 2 phase terms obtained for each discrete frequency, the two frames described above are It is possible to predict the phase of another different frame. Specifically,
When the data of the phase term of a certain specific frequency value fa in frame 1 is θ1 and the data of the phase term of a certain specific frequency value fa in frame 2 is θ2 as in the above example, the above-mentioned frame Data θ3 of the phase term of a specific frequency value fa in the next frame of 2 is
.theta.3.apprxeq..theta.2 + (. theta.2-.theta.1) = 2.theta.2-.theta.1 (a) is predicted, and the above-described phase prediction is individually performed for all discrete frequencies in frames 1 and 2. It is said that it is possible to predict the phase of the signal of frame 3 by performing the above.

According to the above-mentioned Takahashi's hypothesis, even if the phase information of one frame, for example, the phase information of only frame 1 is known, it is impossible to predict the phase information of other frames using the phase information. However, if the phase information of two frames is known, it is possible to predict the phase information of another frame, and the phase information of two adjacent frames, for example, the phase information of frame 1 and the phase information of frame 2 , It is possible to predict the phase information of frames other than the above-mentioned two frames, and the distance is K times the time length of one frame.
If the phase information of one frame, for example frame 1 and the phase information of frame 4, is known, the phase information of another frame, for example, frame 7, which is K times the time length of one frame away from frame 4, can be predicted. Generally speaking, the above Takahashi's hypothesis can be expressed as follows: "The first and the first in each successive Fourier transform frame cut out from the acoustic signal so as to have a predetermined constant time length.
Discrete Fourier transform is performed on the second Fourier transform frames using the same window function, and the same predetermined number of discrete values obtained as a result of the Fourier transform for each of the first and second Fourier transform frames described above. Phase information for each discrete frequency is obtained for each of the first and second Fourier transform frames described above from the data for each frequency, and the phase information for each of the first and second Fourier transform frames described above is obtained. The phase of change of the phase information for each identical discrete frequency is obtained, and it is assumed that the mode of change of the phase information for each discrete frequency is constant on the time axis and the first time position and the The individual phase information of the predetermined number of discrete frequencies in the Fourier transform frame of the third time position existing at the time position that is an integer multiple of the time difference from the second time position is determined, and Fourier It can be to be able to predict the phase information of the conversion frame ".

The block diagram shown in FIG. 4 is an application of the phase prediction technique for acoustic signals based on the Takahashi's hypothesis as described above.
When the information amount of the signal to be recorded and transmitted is compressed and then recorded and transmitted, the amplitude residual signal Ai (m) -Ai (m-1) and the phase residual signal are calculated for each frame. Difference signal Δθi
(m) = θi (m)-{2θi (m-1) -θi (m-2)} is shown as an example of the configuration of the encoder on the transmission side (encoding side) configured to record and transmit FIG. 5 is a diagram showing a configuration example of a receiving side (decoding side) decoder. Each of the above residual signals becomes zero if the prediction is correct, but usually a slight deviation from the predicted value occurs, so the above residual signals rarely become zero but the original signal The information amount of the residual signal is much smaller than the information amount.

In FIG. 4, reference numeral 1 denotes an input terminal of a digital audio signal to be recorded and transmitted, which is overlapped by a block 2 from the digital audio signal supplied to the input terminal 1 of the digital audio signal. Sequential Fourier transform frames that are cut out so as to have a predetermined fixed time length in the selected state are multiplied by a window function for each period having, for example, N sampling points, and each successive frame The fast Fourier transform operation (FFT operation) is performed in the block 4 after the window function is multiplied in the block 3 so as to form one sequential frame period in which the joints are overlapped with each other and gently connected. Be done. As a result of the FFT calculation, the same constant frequency interval f {however, 1 for each Fourier transform frame is obtained.
Number of data in Fourier transform frame for each frame The number of samples is N, the sampling frequency is fs, and f = fs /
For each of the N discrete frequencies having N}, the real part (Rea
l) The data of the FFT operation result including the amplitude and the imaginary part (Imag) amplitude is obtained.

As described above, the N discrete data for each discrete frequency obtained as a result of the FFT operation is subjected to signal processing by different signal processing devices for each discrete frequency data. However, in FIG. 4, only the configuration of one of the N signal processing devices is illustrated. In FIG. 4, the signal processing device is a component between the block 6 and the multiplexer 18, which is displayed as a rectangular coordinate → polar coordinate conversion.
Data of a specific discrete frequency, which is obtained as a result of the FFT operation and is composed of a real number part and an imaginary number part for each of N discrete frequencies, is polar coordinate-transformed by the orthogonal coordinate → polar coordinate transforming part 6 and becomes an amplitude term. After the separation into the phase term, the calculation of the amplitude according to the above-mentioned Expression 1 and the calculation of the phase according to the above-mentioned Expression 2 are performed, so that the discrete each of the above-described discrete frames is obtained. Composite amplitude term Ai (m) for each frequency
And the phase term θi (m) are obtained.

The composite amplitude term Ai (m) of a specific discrete frequency obtained as the calculation result of the orthogonal coordinate → polar coordinate conversion unit 6 is supplied to the latch circuit 7, the subtractor 8 and the data selector 9. The data selector 9 selects one of the setting data based on the composite amplitude term Ai (m) and the residual data output from the subtracter 8 according to the switching signal supplied to the terminal 34. And outputs it to the multiplexer 18. The switching operation of the data selector 9 described above is performed in conjunction with the switching operation of the data selector 16 described later. Further, the phase term θi (m) of a specific discrete frequency obtained as the calculation result of the above-mentioned rectangular coordinate → polar coordinate conversion unit 6 is determined by the latch circuit 10 and the subtractor 14.
And the data selector 16 are supplied. As a calculation result of the phase of a specific discrete frequency (which is now assumed to be fa) in the m-th frame, the rectangular coordinate → polar coordinate conversion unit 6
The data of the phase θi (m) output from the latch circuit 1
The data of the phase held in the latch circuit 10 before being held at 0, that is, the calculation result of the phase of the specific discrete frequency fa in the (m-1) th frame is output from the orthogonal coordinate → polar coordinate conversion unit 6. Phase θi (m-1)
Is the latch circuit 12 in the phase prediction unit PFC.
Held in. The phase predicting unit PFC is composed of a latch circuit 12, an amplifier 11 having a gain of 2, and a subtractor 13 in the illustrated configuration example.

The phase data held in the latch circuit 12 before the data of the phase θi (m-1) is held in the latch circuit 12, that is, a specific discrete data in the m-2th frame. As a calculation result of the phase of the frequency fa, the phase θ output from the rectangular coordinate → polar coordinate conversion unit 6
The i (m-2) data is supplied to the subtractor 13 as a subtraction signal, and the subtraction signal 13 is supplied to the subtractor 13 as a subtracted signal. From the subtractor 13, that is, the phase prediction unit PF.
The predicted phase data output from C is 2θi (m-1)-
θi (m-2). Phase θ of a specific discrete frequency obtained as the calculation result of the rectangular coordinate → polar coordinate conversion unit 6
i (m) is supplied to the latch circuit 10, the subtractor 13, and the data selector 16. The data of the phase θi (m) output from the rectangular coordinate → polar coordinate conversion unit 6 is held in the latch circuit 10 as the calculation result of the phase of a specific discrete frequency (probably fa) in the m-th frame. As the calculation result of the phase data held in the latch circuit 10 before the operation, that is, the calculation result of the phase of the specific discrete frequency fa in the (m-1) th frame, the orthogonal coordinate → polar coordinate conversion unit 6
The data of the phase θi (m−1) output from the above is held in the latch circuit 12 in the phase predicting unit PFC configured by the latch circuit 12, the amplifier 11 having a gain of 2, and the subtractor 13. The terminal 5 is a shift clock signal supply terminal.

The phase data held in the latch circuit 12 before the data of the phase θi (m-1) is held in the latch circuit 12, that is, a specific discrete data in the m-2th frame. As a calculation result of the phase of the frequency fa, the phase θ output from the rectangular coordinate → polar coordinate conversion unit 6
The data i (m-2) is supplied to the subtractor 13 as a subtraction signal, and the subtraction signal 13 is supplied to the subtractor 13 as a subtracted signal. From the subtractor 13, that is, the phase prediction unit PF.
The predicted phase data output from C is 2θi (m-1)-
θi (m-2). Predicted phase data 2θi (m-1) -θi (m-2) output from the phase predictor PFC described above
Is subtracted from the actual phase data θi (m) in the subtractor 14, so that the phase residual signal Δθi (m) = θi (m)-{2θi (m-1) − from the subtractor 14. θ
i (m−2)} is output and supplied to the quantizing scaling 17 via the data selector 16, where the quantizing size is set according to the frequency, and then supplied to the multiplexer 18.

In the above-mentioned multiplexer 18, the amplitude residual signal ΔAi (m) of a specific discrete frequency (probably fa now) and the phase residual signal Δθi (m) of a specific discrete frequency fa. And are supplied to the output terminal 19.
At the output terminal 19, the original amplitude component Ai (m), phase component θi (m), and
Needless to say, it is supplied. The multiplexer 18 includes all other signal processing circuits that generate the amplitude residual signal ΔAi (m) and the phase residual signal Δθi (m) for each predetermined number of discrete frequencies in the Fourier transform frame. Since the output data from the audio signal is also supplied from the multiplexer 18, the data of the acoustic signal in which the information amount is compressed is output from the multiplexer 18 and sent to the transmission path via the output terminal 19.

Next, an example of the configuration of the receiving side (decoding side) of the transmission system will be described with reference to FIG. In FIG.
20 is an input terminal of a decoder provided on the receiving side,
Data of the acoustic signal in a state in which the amount of information transmitted from the transmitting side to the receiving side via the transmission line (not shown) described above with reference to FIG. 4 is compressed is input to the input terminal 20. That is, a demultiplexer (not shown) is used from the data of the acoustic signal configured to include the amplitude residual signal ΔAi (m) for each predetermined number of discrete frequencies in the Fourier transform frame and the phase residual signal. A signal including the amplitude residual signal ΔAi (m) and the phase residual signal Δθi (m) for each specific discrete frequency separated by is supplied.
The amplitude residual signal ΔAi (m) and the phase residual signal Δθi (m) supplied to the input terminal 20 for each specific discrete frequency.
A signal including and is subjected to predetermined signal processing by a signal processing device that performs signal processing on data of a specific discrete frequency. FIG. 5 typically shows one signal processing device that performs signal processing on data of a specific discrete frequency.

The input terminal 20 includes an original amplitude component Ai (m) of a specific discrete frequency, a phase component θi (m), an amplitude residual signal ΔAi (m) and a phase residual signal Δθi (m). In the signal processing circuit supplied with the signal composed of, the de-multiplexer 21 sets the original amplitude component A of a specific discrete frequency (probably fa).
i (m), phase component θi (m), and amplitude residual signal ΔA
i (m) and the phase residual signal Δθi (m) are separated and supplied to the amplitude component signal processing circuit and the phase component processing circuit. The data given from the latch circuit 24 to the adder 25 in FIG.
To the adder 25 from the amplitude residual signal ΔAi
(m) is the amplitude residual signal ΔAi in the m-th frame
In the case of (m), since it is the data Ai (m-1) of the combined amplitude term of the m-1th frame, m in the adder 25.
-Data Ai (m-1) of the composite amplitude term of the -1st frame,
The amplitude residual signal ΔAi (m) in the mth frame is added, and the adder 25 outputs data Ai (m) of the composite amplitude term of the mth frame, which is held by the latch circuit 24. Data selector 2
It is supplied to the polar coordinate → rectangular coordinate conversion unit 27 via 6.

Also, as described above, the demultiplexer 2
Specific discrete frequencies separated by 1 (for example, fa
The original amplitude component Ai (m), the phase component θi (m), and the phase component θi (m) and the phase residual in the amplitude residual signal ΔAi (m) and the phase residual signal Δθi (m). The difference signal Δθi (m) is requantized by the requantizer 22 and then supplied to the adder 29 and the data selector 30. The fact that the data output from the adder 29 becomes the data θi (m) of the phase term of the specific discrete frequency in the m-th frame means that the two data added by the adder 29 are , The phase residual signal Δ given to the adder 29 from the demultiplexer 21
θi (m) is the phase residual signal Δ in the m-th frame
θi (m) and 2θi output from the phase prediction unit PFC
This is because (m-1) -θi (m-2). The data θi (m) of the phase term of the specific discrete frequency in the m-th frame output from the adder 29 is supplied to the polar coordinate → rectangular coordinate conversion unit 27 via the data selector 30, and It is supplied to the latch circuit 28. A shift clock signal is supplied to the terminal 23.

The phase θ output from the adder 29 described above.
Before the data of i (m) is held in the latch circuit 28,
The phase data held in the latch circuit 28, that is, the specific discrete frequency f in the (m-1) th frame
The data of the phase θi (m-1) of a is held in the latch circuit 12 in the phase predicting unit PFC configured by the latch circuit 12, the amplifier 11 having a gain of 2, and the subtractor 13. The data of the above phase θi (m-1) is the latch circuit 1.
The data of the phase held in the latch circuit 12 before being held in 2, that is, the data of the phase θi (m-2) of the specific discrete frequency fa in the m-2th frame is subtracted by the subtractor 13 Is supplied as a subtraction signal to the subtractor 13, and the subtracted signal is supplied to the subtractor 13 because the output from the amplifier 11 having a gain of 2 is used. The predicted phase data output, that is, the predicted phase data θi (m) output from the phase prediction unit PFC is 2θi (m−1) −θi (m
-2). Therefore, the predicted phase data θi (m) = 2θi (m−1) − output from the phase predictor PFC described above.
θi (m-2) and the phase residual signal Δθi of the m-th frame
When (m) = θi (m)-{2θi (m-1) -θi (m-2)} is added by the adder 29, a specific discrete value in the m-th frame is output from the adder 29. The data θi (m) of the phase term of the frequency fa is output.

M output from the adder 25 as described above
The data Ai (m) of the composite amplitude term of the th frame and the data θi (m) of the phase term of a specific discrete frequency in the mth frame output from the adder 29 are respectively predetermined data selectors. By being supplied to the polar coordinate → rectangular coordinate conversion unit 27 via 26 and 30, the polar coordinate → rectangular coordinate conversion unit 27 causes the data Ai (m) of the composite amplitude term of the m-th frame described above and the adder 29. From the data θi (m) of the phase term of the specific discrete frequency in the m-th frame output from, the real part (Real) amplitude and the imaginary part (Imag) at the specific discrete frequency fa described above. The amplitude and the calculated value are output and supplied to the inverse FFT operation unit 31. Since the inverse FFT operation unit 31 is supplied with output data from all the signal processing circuits provided for each predetermined number of discrete frequencies in the Fourier transform frame, the inverse FFT operation unit 31 also outputs the output data. And the tone signal data are restored, and the window function multiplication shown in block 32 and the overlap addition shown in block 33 are performed to restore the original digital tone signal. Output terminal 35
Will be sent to.

[0024]

By the way, as described above, the data obtained by digitizing the musical tone signal generated from the actual musical instrument or the like is stored in the memory, and the memory storing the data is stored in the memory. In an electronic musical instrument having a configuration such as that provided in a sound source for generating musical tones, and other various devices, if the musical tone signals to be stored in the memory have substantially uniform amplitudes, the memory The amount of data of the amplitude component in the musical tone signal to be stored in can be extremely small, and regarding the phase component in the musical tone signal, the above-mentioned “Phase prediction method of acoustic signal” of the applicant company described above can be used. Thus, “the same window is used for each of the first and second Fourier transform frames in each successive Fourier transform frame cut out from the acoustic signal so as to have a predetermined constant time length. The number of discrete frequencies is discretely Fourier-transformed using a number, and the first predetermined number of data is obtained by the same predetermined number of discrete frequencies obtained as a result of the Fourier transformation of each of the first and second Fourier transform frames. , The phase information for each discrete frequency is obtained for each second Fourier transform frame, and the phase information for each identical discrete frequency corresponding to each other in the first and second Fourier transform frames described above is obtained. Of the phase information of each discrete frequency is constant on the time axis,
That is, assuming that it is in the form of an equal difference term, a predetermined number in the Fourier transform frame of the third time position existing at a time position that is an integral multiple of the time difference between the first time position and the second time position described above. By applying "means for determining individual phase information of discrete frequency and predicting phase information of Fourier transform frame at third time position" to compression of data amount of phase component in tone signal. Can be made smaller,
Therefore, it is possible to facilitate the configuration of the sound source for generating the musical sound in the memory storing the above data.

However, data obtained by digitizing a musical tone signal generated from an actual musical instrument or the like is stored in a memory, and the memory storing the above data is provided in a sound source for generating a musical tone. When the initial setting data of the amplitude component, the initial setting data of the phase component and the value of the equal difference term are used as the data of the tone signal to be stored in the memory in the electronic musical instrument of the configuration form as described above. , When the tone signal is restored using the tone signal data as described above read from the memory, the tone signal with a constant amplitude should originally be synthesized, but in reality the phase component shift amount is accumulated. As a result, the musical tone signal actually synthesized is in a state in which the envelope curve is wavy with the passage of time, that is, on the time axis, as in the waveform illustrated in FIG. 3B. In a state where Oite amplitude is changing. Therefore, there has been a demand for a means capable of preventing the amplitude of the tone signal from changing in the time axis without increasing the data amount of the tone signal stored in the memory.

[0026]

SUMMARY OF THE INVENTION The present invention discretely Fourier transforms a signal of each successive Fourier transform frame cut out from a musical tone signal so as to have a predetermined fixed time length by using the same window function. A high-efficiency code is obtained by transforming the amplitude component and the phase component of each discrete frequency obtained by using the same predetermined number of data of each discrete frequency obtained as a result of the Fourier transform for each Fourier transform frame described above. In the method of synthesizing a tone signal such that the tone signal is synthesized by performing inverse Fourier transform after dequantization, for each phase component for each discrete frequency obtained for each Fourier transform frame described above, M-1 cut out from the tone signal so as to have a predetermined fixed time length
When the phase of the specific discrete frequency in the th Fourier transform frame is θm−1 and the phase of the specific discrete frequency in the m−2th Fourier transform frame is θm−2, the phase in the mth Fourier transform frame is The phase θm of a specific discrete frequency is θm = θm-1 + (θm-1−
When performing high-efficiency encoding by predicting as follows: θm-2) = θm-1 + θa, the value θa of the phase difference component is rounded to a value of 2π / n (where n is an integer). Provided is a method of synthesizing a musical tone signal, which is capable of generating a continuous sound with a small amount of data only by setting a parameter in a frequency domain and changing a period of the continuous sound.

[0027]

The signal of each successive Fourier transform frame cut out from the tone signal so as to have a predetermined fixed time length is discretely Fourier transformed using the same window function.
An amplitude component and a phase component for each discrete frequency obtained by using the same predetermined number of data for each discrete frequency obtained as a result of the Fourier transform for each Fourier transform frame described above are obtained. The initial setting data of the amplitude component and the initial setting data of the phase component are stored in the memory. Further, the phase component for each discrete frequency obtained for each Fourier transform frame is specified in the m-1th Fourier transform frame cut out from the musical tone signal so as to have a predetermined constant time length. The phase of the discrete frequency is θm−1, and m
-When the phase of the specific discrete frequency in the -2nd Fourier transform frame is θm-2, the phase θm of the specific discrete frequency in the m-th Fourier transform frame is
θm = θm-1 + (θm-1−θm-2) = θm-1 + θa The value θa of the difference term is rounded to a value of 2π / n (where n is an integer) and then stored in memory. .

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The specific contents of the tone signal synthesizing method of the present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 is a block diagram showing a configuration example of an encoder that can be used when implementing the tone signal synthesizing method of the present invention, and FIG. 2 is a block diagram showing an example configuration of a decoder that can be used when implementing the tone signal synthesizing method of the present invention. FIG. 3 is a diagram used for explaining the operation, FIG. 4 is a block diagram showing a configuration example of an encoder used in the already-proposed acoustic signal transmission method, and FIG. 5 is used in the already-proposed acoustic signal transmission method. FIG. 6 is a block diagram showing a configuration example of a decoder to be used, and FIG. 6 is a diagram for explaining a configuration principle and an operation of a proposed transmission method of an acoustic signal. 1 shows an example of the configuration of an encoder that can be used in the implementation of the tone signal synthesizing method of the present invention. The components corresponding to those in the encoder described with reference to FIG. 4 are shown in FIG. The same reference numerals as those used in the encoder are used, and in FIG. 2 showing an example of the configuration of a decoder that can be used in implementing the musical tone signal synthesizing method of the present invention, it has already been described with reference to FIG. The same reference numerals as those used in the decoder shown in FIG. 5 are used for the components corresponding to those in the decoder.

In FIG. 1, which shows an example of the configuration of an encoder used in the method of synthesizing a tone signal of the present invention, reference numeral 1 denotes a digital tone signal (digital tone signal obtained by converting a tone signal generated from an actual musical instrument or the like into a digital signal. Audio signal) input terminal, which is cut out from the digital musical tone signal supplied to the digital musical tone signal input terminal 1 described above so as to have a predetermined time length in an overlapped state by the block 2. The generated sequential Fourier transform frames are multiplied by a window function for each period having, for example, N sampling points, so that the joints of the successive frames are overlapped with each other and gradually connected. After being multiplied by the window function in block 3 so as to be one frame period, in block 4, the fast Fourier transform operation (FF
T calculation) is performed. Then, as a result of the FFT operation in the block 4 described above, the same constant frequency interval f {however, 1 for each Fourier transform frame is obtained.
Number of data in Fourier transform frame for each frame The number of samples is N, the sampling frequency is fs, and f = fs /
A real part (Rea) for every N discrete frequencies having N}.
l) FFT consisting of amplitude and imaginary part (Imag) amplitude
Data of the calculation result is obtained.

As described above, the N discrete frequency data obtained as a result of the FFT operation is subjected to signal processing by different signal processing devices for each discrete frequency data. However, in FIG. 1, only the configuration of one signal processing device among the N signal processing devices is shown as a block 6 shown by a dot-dash line, which is displayed as Cartesian coordinates → Polar coordinates conversion in the drawing. It is shown between the multiplexer 18. Data of a specific discrete frequency, which is obtained as a result of the FFT operation and is composed of a real number part and an imaginary number part for each of N discrete frequencies, is polar coordinate-transformed by the orthogonal coordinate → polar coordinate transforming part 6 and becomes an amplitude term. After being separated into the phase term, the amplitude calculation by the above-mentioned equation 1 is performed by the amplitude calculator 6A, and the phase calculation by the above-mentioned equation 2 is performed by the phase calculator 6P, so that For each discrete frequency described above for the frame, the amplitude component Ai
(M) and the phase component θi (m) are obtained. Then, the above-described Cartesian coordinate → polar coordinate conversion unit 6 has an amplitude calculation unit 6
The amplitude component Ai (m) of the specific discrete frequency obtained as the calculation result in A is supplied to the masking processing unit 15,
In addition, the phase component θi (m) of the specific discrete frequency obtained as the calculation result of the phase calculation unit 6P in the Cartesian coordinate → polar coordinate conversion unit 6 is the generation unit 3 of the difference term data θa.
6 and the fixed contact a of the changeover switch 40.

Amplitude calculation section 6 of Cartesian coordinate to polar coordinate conversion section 6
The masking processing unit 1 supplied with the amplitude component Ai (m) of a specific discrete frequency obtained as the calculation result in A
In 5, masking processing is performed on the amplitude components in the supplied sequential Fourier transform frames. For example, the masking curve mask (m) is calculated by using the amplitude component Ai (m) of the specific discrete frequency described above, and then the human auditory perception characteristic that the human auditory sense is low in the low range and the high range. The final masking curve mask (m) is determined using the so-called loudness curve corresponding to Then, in the masking processing unit 15, the amplitude component Ai (m) is masked by the masking curve mask (m) determined as described above, that is, Ai (m).
Ai (m) in the case of (m)> mask (m) outputs it as it is, and Ai (m) in the case of Ai (m) <mask (m) is set to zero. Then, the data amount of the amplitude component is compressed.

The amplitude component Ai (m) whose data amount has been compressed by the masking processing section 15 as described above is supplied to the fixed contact a of the changeover switch 38. The movable contact v of the changeover switch 38 described above is changed by the changeover signal supplied to the terminal 39 to the orthogonal coordinate → polar coordinate conversion unit 6 described above.
Among the amplitude components Ai (m) of the specific discrete frequency obtained as the calculation result in the amplitude calculation unit 6A, the data of the amplitude component in the Fourier transform frame to be used as the initial setting data is set as the initial setting data. It is supplied to the adaptive quantizer 42 at a predetermined timing. In the adaptive quantization unit 42 described above, the adaptive quantization processing is performed in the following procedure, for example. That is, the adaptive quantizer 42 first detects the maximum absolute value max (k) of the amplitude component Ai (m) supplied thereto for each band, and then detects the amplitude component detected for each band as described above. The required number of bits bit (k) is determined for each maximum value max (k) of the absolute value of Ai (m).

Then, the above-mentioned max (k), bit (k)
On the basis of the above, the amplitude component Ai (m) which is not in the zero state is adaptively quantized, the data existence flag is set, and the initialization data Ai ′ of the adaptively quantized amplitude component is set.
(m), that is, Ai ′ (m) = {Ai (m) / max
(k)} is supplied to the multiplexer 18. The changeover operation of the changeover switch 38 described above is performed by the changeover switch 4 described later.
This is performed in conjunction with the 0 switching operation. Further, the movable contact v of the changeover switch 38 is changed over to the fixed contact b side of the changeover switch 38 except when the data of the amplitude component in the Fourier transform frame in which the amplitude component is to be used as the initial setting data is transmitted. However, no data is supplied to the adaptive quantizer 42.

In addition to the adaptive quantization performed by the adaptive quantizer 42, the masking curve mas
Amplitude component Ai (m) masked by k (m), that is, amplitude component Ai (m) <masking curve mask
Since the amplitude component Ai (m) having the relationship of (m) cannot be heard, it is not necessary to transmit the data of such amplitude component, and the flag (flag1, flag indicating whether or not the data exists). send flag0). Also,
An amplitude component Ai (m) that is not masked by the masking curve mask (m), that is, an amplitude component A having a relationship of amplitude component Ai (m)> masking curve mask (m).
It is not necessary to send all data for i (m) either, and the maximum value max (k) (k is 0 to 3 to 6 for each frequency band).
The number of bits required for each frequency band based on the maximum value max (k) detected for each frequency band, bit (k) {bit (k), for example, is about 14 bits at maximum. } Should be allocated. Then, in the adaptive quantizer 42 described above, it is necessary for each frequency band based on the maximum value max (k) detected for each frequency band and the maximum value max (k) detected for each frequency band. The number of bits bit (k) is used to quantize the amplitude component in which data exists according to the state of the flag, and the adaptively quantized amplitude component data Ai ′ (m), that is, Ai ′ (m).
= {Ai (m) / max (k)} is calculated and supplied to the multiplexer 18.

The data of the amplitude component supplied to the multiplexer 18 and sent from the multiplexer 18 to the terminal 19 is in a state of being highly efficient coded in both the direction of the frequency axis and the direction of the time axis. As compared with the case of performing high-efficiency encoding by the conventional high-efficiency encoding method, the data amount is significantly compressed. As described above, the reason why the data amount can be significantly reduced compared to the case where the high efficiency encoding is performed by the conventional high efficiency encoding method is that the music signal is cut out so as to have a predetermined constant time length. Discrete Fourier transform using the same window function on the signals of each successive Fourier transform frame,
An amplitude component and a phase component for each discrete frequency obtained for each Fourier transform frame using the same predetermined number of discrete frequency data obtained as a result of the Fourier transform for each Fourier transform frame. Are individually transmitted with high efficiency and then transmitted.

Next, the phase component θi (m) of the specific discrete frequency obtained as the calculation result of the phase calculation unit 6P in the rectangular coordinate → polar coordinate conversion unit 6 is the difference term data generation unit 3
6 and the fixed contact a of the changeover switch 40. The differential term data generation unit 36 is provided with a memory, and with respect to the phase component for each specific discrete frequency in the sequential Fourier transform frame stored therein, for example, the specific discrete frequency in the m-1st Fourier transform frame Each phase is θi (m-1), and the phase for each specific discrete frequency in the m-2nd Fourier transform frame is θi (m-2).
, The phase for each specific discrete frequency in the m-th Fourier transform frame is θi (m), θi (m) = θi
(m-1) + {θi (m-1) -θi (m-2)} = θi (m-1) + θi
It is configured so as to have a function of calculating the value θia of the difference term when the prediction is performed as shown by a.

The difference term value θia generated by the difference term data generator 36 is the difference term value θi described next.
It is supplied to the correction part 37 for the data of the difference term, which has a function of rounding a to 2π / n (where n is a positive integer).
In the above-mentioned equalization term correction unit 37, the value θia of the equality term generated by the above-mentioned equality term data generation unit 36 is set to 2π.
/ N (where n is a positive integer) is output as data rounded and supplied to the fixed contact b of the changeover switch 40. The operation of the above-described correction item 37 for the difference term data will be described below with reference to a specific example. now,
2π / n set in the above-mentioned equality term correction unit 37
Is, for example, 2π / 16, and the value of the differential term supplied from the differential term data generation unit 36 to the differential term correction unit 37 is, for example, 46 degrees. In the correction term 37 of the differential term, the value closest to the value of the differential term of 46 degrees supplied thereto is 2π / 16 = 36.
An angle that is an integral multiple of 0/16 = 22.5 degrees, that is, 2
2.5 degrees, 45 degrees, 67.5 degrees, 90 degrees ... 360 degrees, and find a value corresponding to that multiple (in the above example, the angle 45 degrees of the multiple 2 is the closest to 46 degrees). 2) is supplied to the multiplex 18 instead of the numerical value of the angle 46 described above.

As described above, the change-over switch 40 is supplied with the change-over control signal to the control signal supply terminal 41 so as to perform the change-over operation in synchronization with the change-over switch 38. The movable contact v of the changeover switch 40 is switched to the fixed contact a side by the changeover control signal supplied to the terminal 41 only when transmitting the data of the phase component in the Fourier transform frame to be transmitted first. The initial setting value data of the phase component, for example, the phase component data θi (0) output from the phase calculation unit 6P at this time is the fixed contact a and the movable contact v of the changeover switch 40. Is supplied to the adaptive quantizing unit 43 via. In the adaptive quantizer 43, the data of the phase component supplied via the movable contact v of the changeover switch 40 is adaptively quantized in the multiplexer 18 by referring to, for example, the look-up table LUT. Supply. As the above-mentioned look-up table LUT, a look-up table LUT in which data of the number of bits in which the required number of bits is determined by using the signal level (amplitude) value and the frequency value as parameters can be used can be used. By giving the information of the signal level (amplitude) value and the frequency value to the table LUT as the address signal from the amplitude calculating section 6A, the data of the bit number corresponding to the address is given to the adaptive quantizing section 43. Thus, the phase component supplied to the adaptive quantizer 43 via the fixed contact a and the movable contact v of the changeover switch 40 corresponds to the number of bits read from the look-up table LUT. The data is converted into a number of data and output from the adaptive quantization unit 43.

Therefore, the data of the musical tone signal transmitted from the multiplexer 18 in the encoder shown in FIG. 1 through the output terminal 19 is the Fourier transform frame whose amplitude component should be used as the initial setting data as described above. In the initial setting data of the amplitude component, which is highly efficient quantized by performing the masking process and the adaptive quantization on the amplitude component in, and the data amount is largely compressed, the amplitude component is set as the initial setting data. Fourier transform of the phase component to be used Adaptive quantization is applied to the phase component in the frame, and the initial setting data of the phase component that is highly efficient quantized and is in a state in which the amount of data is greatly compressed, and the sequential Fourier transform The phase component of each specific discrete frequency in the frame is, for example, the m-1st Fourier transform frame If the phase for each specific discrete frequency is θi (m-1) and the phase for each specific discrete frequency in the m-2nd Fourier transform frame is θi (m-1), then the m-th The phase θi (m) for each specific discrete frequency in the Fourier transform frame of
With respect to the value θia of the differential term in the prediction such that i (m-1) + {θi (m-1) −θi (m-2)} = (m-1) + θia, θia
Is rounded to 2π / n (where n is a positive integer), and is obtained, and the musical tone signal is quantized with high efficiency. The data is in a state of being largely compressed as compared with the case where the data amount is compressed. The data of the musical tone signal is output from the output terminal 19 of the encoder and stored in the memory of the tone generator of the electronic musical instrument.

In the tone signal data as described above, the value θia of the equal difference term is 2π / when the phase prediction is performed as described above for the phase component for each specific discrete frequency in the successive Fourier transform frames. n (however, n
Is rounded to a positive integer). Therefore, when the tone signal data described above is read out from the memory of the electronic musical instrument and is decoded by the decoder described later to be synthesized as a tone signal, the tone signal has a cycle of n frames. Since it becomes a repetitive signal, like the waveform of the musical tone signal illustrated in FIG.
The envelope does not wavy over time.

Next, the tone signal data read from the memory of the tone generator of the electronic musical instrument, that is, the amplitude component in the Fourier transform frame where the amplitude component should be used as the initial setting data as described above, is highly efficient quantized. The initial setting data of the amplitude component in a state in which the data amount is compressed significantly and the phase component in the Fourier transform frame in which the amplitude component should be used as the initial setting data are highly efficiently quantized and the data amount is compressed significantly. Regarding the initial setting data of the phase component of the state and the value θia of the phase difference of the phase component for each specific discrete frequency in the sequential Fourier transform frame, θia is 2π / n (where n is a positive value). 2) for the configuration of an example of a decoder that is used when decoding the numerical data obtained by rounding to the It described Te.

In FIG. 2, reference numeral 20 denotes an input terminal of the signal processing circuit for a signal of a specific discrete frequency (probably fa) among a large number of signal processing circuits provided in the decoder. Another signal processing circuit having the same configuration as that of the signal processing circuit shown in FIG. 2 (a signal processing circuit for processing a signal having a specific discrete frequency other than fa) is shown in FIG. Is omitted. The specific discrete frequency (currently f
The signal of the amplitude component and the signal of the phase component of (a) are separated from the signal of the amplitude component and the signal of the phase component described above.
The demultiplexer 4 demultiplexes the initial setting data of the amplitude component of the specific discrete frequency separated by the multiplexer 21.
4, the initial setting data of the phase component of the specific discrete frequency is supplied to the inverse quantizer 45, and the differential component data of the phase component is supplied to the decoder 46.

The dequantizer 44 dequantizes the initial setting data of the amplitude component read from the memory of the sound source of the electronic musical instrument, and the output data is latched by the latch circuit 47. The output data from the above-mentioned latch circuit 47 is supplied to the polar coordinate → orthogonal coordinate transforming section 27 and the look
It is also given to the up table 53. The dequantizer 45 dequantizes the initial setting data of the phase component read from the memory of the sound source of the electronic musical instrument.
That is, in the inverse quantizer 45 to which the necessary number of bits of information is given from the look-up table 53 to which the information of the magnitude of the amplitude component of the specific discrete frequency is supplied as described above, The initial setting data of the phase component is inversely quantized in the range of 0 to 2π based on the flag, and the output data is supplied to the fixed contact a of the changeover switch 49. Therefore, when the movable contact v of the changeover switch 49 is changed over to the fixed contact a side, the output data from the inverse quantizer 45 is the polar coordinate → orthogonal coordinate conversion unit 2.
7 and is latched by the latch circuit 51.

Further, the phase difference component data (numerical value) separated by the demultiplexer 21 and supplied to the decoder 46 is the value of the angle θia of the phase component difference term by the decoder 46. And is supplied to the latch circuit 48. The output data from the latch circuit 48 and the output data from the latch circuit 51 are added by the adder 52, so that the adder 52 outputs a phase component in successive Fourier transform frames, and the phase component is switched. It is supplied to the polar coordinate → rectangular coordinate conversion unit 27 via the fixed contact b and the movable contact v of the switch 49.

The polar coordinates → orthogonal coordinates conversion unit 27 uses the data of the amplitude component of the Fourier transform frame described above and the data of the phase component of a specific discrete frequency in the Fourier transform frame output from the adder 52. , The real part at the specific discrete frequency fa described above
The (Real) amplitude and the imaginary part (Imag) amplitude are calculated and output, and are supplied to the inverse FFT operation unit 31. The inverse FFT operation unit 31 is supplied with output data from all the signal processing circuits provided for each predetermined number of discrete frequencies in the Fourier transform frame. The original musical sound signal is restored. . And the above-mentioned reverse F
The tone signal output from the FT calculator 31 is subjected to window function multiplication shown in block 32 and overlap addition shown in block 33,
The original tone signal is restored and output to the output terminal 35.

[0046]

As is apparent from the above detailed description, the method of synthesizing a tone signal of the present invention is such that each forward Fourier transform cut out from the tone signal so as to have a predetermined fixed time length. The discrete Fourier transform is performed on the signal of the frame using the same window function, and each discrete transform obtained by using the same predetermined number of discrete frequency data obtained as a result of the Fourier transform for each Fourier transform frame described above. Obtaining the amplitude component and the phase component for each frequency, storing the initial setting data of the amplitude component and the initial setting data of the phase component in the memory, and for each discrete frequency obtained for each Fourier transform frame described above. Regarding the phase component, the phase of the specific discrete frequency in the m-1 th Fourier transform frame cut out from the musical tone signal so as to have a predetermined constant time length is θm−1, When the phase of the specific discrete frequency in the m−2th Fourier transform frame is θm−2, the phase θm of the specific discrete frequency in the mth Fourier transform frame is θm = θm−1 + (θm−1− θ
m-2) = θm-1 + θa, the value of the difference term θa
Is rounded to a value of 2π / n (where n is an integer) and then stored in the memory, so the data obtained by digitizing the musical tone signal generated from the actual musical instrument is stored in the memory. In addition, the amplitude component is used as the data of the tone signal to be stored in the memory of the electronic musical instrument having the configuration in which the memory storing the data is provided in the sound source for generating the tone. Even when the initial setting data of, the initial setting data of the phase component, and the value of the differential term are used, the phenomenon occurs that the envelope curve becomes wavy as time elapses because it becomes a repetitive signal in the cycle of n frames. The musical tone signal having a constant amplitude is synthesized as shown in the waveform of FIG. 3 (a) without being generated, and according to the present invention, the musical tone signal data stored in the memory is recorded. As well as very small amounts, yet it than can be easily obtained those never amplitude changes on the time axis as a reproduced musical tone signals, conventional problems by the present invention are all well resolved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration example of an encoder that can be used in implementing a musical sound signal synthesizing method of the present invention.

FIG. 2 is a block diagram showing an example of the configuration of a decoder that can be used when implementing the musical tone signal synthesizing method of the present invention.

FIG. 3 is a diagram used for explaining an operation.

FIG. 4 is a block diagram showing a configuration example of an encoder used in the already proposed method of transmitting an acoustic signal.

FIG. 5 is a block diagram showing a configuration example of a decoder used in the already proposed method of transmitting an acoustic signal.

FIG. 6 is a diagram for explaining a configuration principle and an operation of a proposed transmission method of an acoustic signal.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input terminal, 6 ... Cartesian coordinate-> polar coordinate conversion part, 6A ... Amplitude calculation part, 6P ... Phase calculation part, 7, 10, 12, 24,
28, 47, 48, 51 ... Latch circuit, 8, 13, 14
... Subtractor, 15 ... Masking processing unit, 38, 40, 49
... Changeover switch, 11 ... Amplifier with gain of 2, 42, 43
... adaptive quantizer, 18 ... Multiplexer, 19 ... Output terminal, 20 ... Decoder input terminal, 21 ... Demultiplexer, 25, 29, 52 ... Adder, 27 ... Polar coordinate → Cartesian coordinate converter, 31 ... Inverse FFT Computation unit, 36 ... Equal difference data generation unit, 37 ... Equal difference data correction unit, 44, 45
... inverse quantizer, 46 ... decoder, 53, LUT ... look
Up table,

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] November 6, 1992

[Procedure Amendment 1]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 1

[Correction method] Change

[Correction content]

[Figure 1] ─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] August 18, 1993

[Procedure Amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0018

[Correction method] Change

[Correction content]

In the above-mentioned multiplexer 18, the amplitude residual signal ΔAi (m) of a specific discrete frequency (probably fa now) and the phase residual signal Δθi (m) of a specific discrete frequency fa. And are supplied to the output terminal 19.
It goes without saying that the output terminal 19 is supplied with the original amplitude component Ai (m) and the phase component θi (m) at least once. The multiplexer 18 includes all other signal processing circuits that generate the amplitude residual signal ΔAi (m) and the phase residual signal Δθi (m) for each predetermined number of discrete frequencies in the Fourier transform frame. Since the output data from the audio signal is also supplied from the multiplexer 18, the data of the acoustic signal in which the information amount is compressed is output from the multiplexer 18 and sent to the transmission path via the output terminal 19.

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0039

[Correction method] Change

[Correction content]

Therefore, the data of the musical tone signal transmitted from the multiplexer 18 in the encoder shown in FIG. 1 through the output terminal 19 is the Fourier transform frame whose amplitude component should be used as the initial setting data as described above. In the initial setting data of the amplitude component, which is highly efficient quantized by performing the masking process and the adaptive quantization on the amplitude component in, and the data amount is largely compressed, the amplitude component is set as the initial setting data. Fourier transform of the phase component to be used Adaptive quantization is applied to the phase component in the frame, and the initial setting data of the phase component that is highly efficient quantized and is in a state in which the amount of data is greatly compressed, and the sequential Fourier transform The phase component of each specific discrete frequency in the frame is, for example, the m-1st Fourier transform frame If the phase for each specific discrete frequency is θi (m-1) and the phase for each specific discrete frequency in the m-2nd Fourier transform frame is θi (m-1), then the m-th The phase θi (m) for each specific discrete frequency in the Fourier transform frame of
i (m-1) + {θi (m-1) -θi (m-2)} = θi (m-1) + θia
Regarding the value θia of the differential term when making a prediction as
and the numerical data obtained by rounding θia to 2π / n (n is a positive integer), and the musical tone signal is highly efficiently quantized, so that the data amount is highly efficient encoded by the conventional method. Thus, the data is in a state of being largely compressed as compared with the case where the data amount is compressed. The data of the musical tone signal is output from the output terminal 19 of the encoder and stored in the memory of the tone generator of the electronic musical instrument.

[Procedure 3]

[Document name to be amended] Statement

[Name of item to be corrected] 0045

[Correction method] Change

[Correction content]

The polar coordinates → orthogonal coordinates conversion unit 27 uses the data of the amplitude component of the Fourier transform frame described above and the data of the phase component of a specific discrete frequency in the Fourier transform frame output from the adder 52. , The real part at the specific discrete frequency fa described above
The (Real) amplitude and the imaginary part (Imag) amplitude are calculated and output, and are supplied to the inverse FFT operation unit 31. The inverse FFT operation unit 31 is supplied with output data from all the signal processing circuits provided for each predetermined number of discrete frequencies in the Fourier transform frame. The original musical sound signal is restored. And the above-mentioned reverse FF
The tone signal output from the T calculation unit 31 includes the block 3
By applying the window function multiplication shown by 2 and the overlap addition shown by block 33, the original tone signal is restored and output to the output terminal 35.

[Procedure amendment 4]

[Document name to be corrected] Drawing

[Name of item to be corrected] Figure 3

[Correction method] Change

[Correction content]

[Figure 3]

Claims (1)

[Claims] 1. The Fourier transform is performed discretely by using the same window function on the signal of each successive Fourier transform frame cut out from the musical tone signal so as to have a predetermined constant time length, and each Fourier transform described above. The high-efficiency encoding of the amplitude component and the phase component of each discrete frequency obtained by using the same predetermined number of data of each discrete frequency obtained as a result of the Fourier transform for each transform frame, and the Fourier component after inverse quantization are performed. In the method of synthesizing a musical tone signal, which is adapted to perform the inverse transformation to synthesize a musical tone signal, the phase component for each discrete frequency obtained for each Fourier transform frame described above is fixed for a predetermined time. The phase of a specific discrete frequency in the m−1th Fourier transform frame cut out from the musical tone signal so as to have a length is θm−1,
When the phase of the specific discrete frequency in the m-2nd Fourier transform frame is θm-2, the phase θm of the specific discrete frequency in the mth Fourier transform frame
Is θm = θm-1 + (θm-1−θm-2) = θm-1 + θa
When high-efficiency encoding is performed by predicting as follows, the value θa of the phase difference component is 2π / n (where n
Is a whole number), the continuous tone can be generated with a small amount of data only by setting the frequency domain parameters, and the period of the continuous tone can be changed. .