JP4330381B2 - Noise removal device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
According to the present invention, a digital signal generated by sequentially sampling an analog signal is divided into blocks each having a predetermined number of sample values, and a result of applying a window to this block and the time axis. The present invention relates to a noise removing apparatus that performs a smoothing process for eliminating discontinuity at a connection point when both blocks are connected to a block separated by a predetermined number of blocks on the rear side.
[0002]
[Prior art]
Conventionally, when a digital signal obtained by digitizing an analog signal as a musical sound signal or the like is processed by being divided into a plurality of blocks by a computer, discontinuity occurs at connection points between adjacent blocks. As a method for eliminating this discontinuity, first, a window function for applying a window before performing fast Fourier transform (FFT) on a plurality of blocks, and then applying a window after performing inverse fast Fourier transform (inverse FFT). There is a method of designing each window function so as to be a triangular window or a Hanning window by multiplying with the above window functions. In addition, a conventional noise reduction system is proposed in which the number of samples subjected to FFT processing in a block is changed so that the causality is maintained even if the digital signal is divided into a plurality of blocks and processed. (For example, refer to Patent Document 1).
[0003]
[Patent Document 1]
Japanese translation of PCT publication No. 2002-517021 (page 13, FIG. 1)
[0004]
[Problems to be solved by the invention]
However, in the method of eliminating the discontinuity by designing the window function as described above, the computer performs processing for applying a window before the fast Fourier transform (FFT) and after the inverse fast Fourier transform (inverse FFT). However, there is a problem that the amount of calculation to be processed becomes enormous.
[0005]
In addition, the method of changing the number of samples to be subjected to FFT processing in a block has a problem that the amount of calculation processed by the computer becomes enormous by changing the number of samples, which cannot be handled by a computer having low calculation capability. It was.
[0006]
Therefore, the present invention has been made in consideration of such circumstances, and as a result of applying a window, when two blocks separated by a predetermined number of blocks that are adjacent on the time axis are connected, those When discontinuity occurs at the connection point between both blocks, the discontinuity is eliminated without increasing the amount of computation while maintaining the original number of samples by correcting the waveforms of the two blocks to approach each other. An object of the present invention is to provide a noise removing device that can perform the above-described operation.
[0007]
[Means for Solving the Problems]
In order to solve the above problems, the present invention provides a certain block (front block) in a block group obtained by dividing a digital signal generated by sequentially sampling analog signals into blocks each having a predetermined number of sample values. As a result of applying a window, the discontinuity at the connection point when both blocks are connected with the block (rear block) separated by a predetermined number of blocks on the rear side that became adjacent on the time axis. A noise removal device that performs smoothing processing to eliminate,
Correction amount calculation means for obtaining a correction amount when performing the smoothing process at a connection point where discontinuity between the two blocks has occurred;
A difference calculating means for dividing the obtained correction amount by a smoothing score indicating the number of sample values to be subjected to the smoothing processing among the sample values in the block to obtain a difference value per sample;
A process of adding a numerical value calculated based on the difference value and the certain variable to a sample value at a position corresponding to a certain variable in the rear block is defined as an addition process, and from the head of the rear block For each of up to the number of sample values indicated by the number of smoothing points, an addition means for sequentially changing the certain variable from the number of times indicated by the number of smoothing points to the top side and repeatedly performing the addition process When,
A process of subtracting a numerical value calculated based on the difference value and the certain variable from a sample value at a position corresponding to the certain variable in the front block is defined as a subtraction process, and from the end of the front block, For each sample value up to the number before the number indicated by the number of smoothing points, the certain variable is sequentially changed from the number before the number indicated by the number of smoothing points to the end side, and the subtraction process is repeated. And a subtracting means.
[0008]
According to such an invention, the adding means sets a certain variable from the number indicated by the smoothing point to the head side for each of the sample values indicated by the smoothing point from the beginning of the rear block. Sequentially change and repeat the addition process, and the subtracting means starts from the last number of smoothing points from the last for each sample value from the end of the previous block to the number of smoothing points. Since a certain variable is sequentially changed toward the end side and the subtraction process is repeated, in the rear block and the front block, up to the number of sample values indicated by the smoothing points from the head of the rear block, the front block The waveform corresponding to the sample value from the end to the number before the number of smoothing points is corrected and connected It is possible to eliminate discontinuity at.
[0009]
In the above-described noise removing device,
The correction amount calculating means includes
Means for obtaining an amplitude difference at a connection point where the discontinuity between the two blocks has occurred, and means for obtaining a correction amount when performing the smoothing process by dividing the obtained amplitude difference by 2 Also good.
[0010]
According to such an invention, the amplitude difference at the connection point where the discontinuity between the two blocks occurs is divided by 2 to obtain the correction amount when performing the smoothing process. Since the waveform of the block is corrected so as to approach, the amplitude difference is eliminated and the discontinuity at the connection point can be eliminated.
[0011]
Furthermore, in the above-described noise removing apparatus, the digital signal may be a musical tone signal. According to such an invention, it is possible to eliminate the discontinuity generated in the digital signal as the musical sound signal.
[0012]
As another noise removing apparatus according to the present invention, a certain block (front block) in a block group obtained by dividing a digital signal generated by sequentially sampling an analog signal into blocks each having a predetermined number of sample values. ) And this and a block that is a predetermined number of blocks away on the time axis (rear side block) is assumed to be connected, and noise that performs smoothing processing to eliminate discontinuity at the connection point A removal device,
Correction amount calculating means for obtaining a correction amount used when performing the smoothing process at a connection point where discontinuity occurs between the two blocks;
Difference calculation means for obtaining a difference value per sample by dividing the obtained correction amount by the number of smoothing points which is the number of points (smoothing points) corresponding to samples belonging to the portion to be smoothed in one block. When,
A process of adding a value obtained by multiplying a variable value that changes according to a smoothing point position and the difference value to a sample value corresponding to the smoothing point position for each of a plurality of smoothing points in the rear block. Addition processing means for performing
A process of subtracting, for each of a plurality of smoothing points in the front block, a sample value corresponding to the smoothing point position, a value obtained by multiplying the variable value that changes according to the smoothing point position and the difference value. And a subtracting means for performing the processing.
[0013]
In the above-described noise removing apparatus,
The addition processing means includes
A value obtained by multiplying the variable J by the difference value with respect to the sample value corresponding to the “smoothing point−J (J is a natural number variable that changes from 1 to maximum smoothing point)” position from the head of the rear block. The process of adding is configured to change the variable J from 1 to the maximum number of smoothing points,
The subtraction processing means includes
A value obtained by multiplying the variable J by the difference value with respect to the sample value corresponding to the position “Smoothing point-J (J is a natural number variable that changes from 1 to the maximum smoothing point)” from the end of the front block. The subtracting process may be performed by changing the variable J from 1 to the maximum number of smoothing points.
[0014]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block configuration diagram of a noise removing device 10 according to the present embodiment. The noise removal apparatus 10 includes a CPU 100, a ROM 110, a RAM 120, an A / D (analog / digital converter) 130, and a D / A (digital / analog converter) 140.
[0015]
The CPU 100 executes a program stored in the ROM 110 to add an effect of an acoustic effect to the musical sound signal converted into a digital signal by the A / D 130 by processing in the frequency domain.
[0016]
Further, the CPU 100 executes a program stored in the ROM 110 to perform a smoothing process, which will be described later, on a digital signal as a musical signal with a window. That is, the rear side of a block group obtained by dividing a digital signal as a musical sound signal into blocks each composed of a predetermined number of sample values and adjacent to this on the time axis as a result of applying a window When both blocks are connected to a block separated by a predetermined number of blocks, smoothing processing is performed to eliminate discontinuities at the connection point between those blocks.
[0017]
For example, as a result of applying a window to a digital signal as a musical sound signal with a half overlap, as shown in FIG. 4, the first block, the third block, the second block, and 4 When the second blocks are adjacent to each other on the time axis, smoothing processing is performed to eliminate the discontinuity that occurs between the first and third blocks and the connection point between the second and fourth blocks.
[0018]
In the ROM 110, processing for applying a window to a digital signal as a musical sound signal, processing for fast Fourier transform (hereinafter referred to as FFT), processing for inverse fast Fourier transform (hereinafter referred to as inverse FFT), frequency domain A program for causing the CPU 100 to execute a process for adding an effect of an acoustic effect or the like by the process in is stored. In addition, a program for causing the CPU 100 to execute the smoothing process is stored.
[0019]
The RAM 120 stores the spectrum of the noise waveform detected in order to remove the noise waveform from the digital signal as the musical sound signal. Also, the amplitude difference at the discontinuous connection point generated at the connection point between the two blocks that are adjacent on the time axis, the correction amount at the time of the smoothing process, and the like calculated by the CPU 100 in the smoothing process are stored. It has come to be.
[0020]
The A / D 130 has a function of generating a digital signal as a musical tone signal by sequentially sampling an analog signal as a musical tone signal received from the outside of the noise removing device 10 on the time axis. In addition, it has a function of inputting the generated digital signal to the CPU 100.
[0021]
The D / A 140 has a function of converting a digital signal as a musical sound signal output from the CPU 100 into an analog signal as a musical sound signal. Further, it has a function of sending out an analog signal as a musical tone signal to the outside of the noise removing device 10.
[0022]
Subsequently, processing in the noise removing apparatus 10 having the above-described configuration will be described with reference to flowcharts shown in FIGS. First, the overall processing in the noise removal apparatus 10 will be described with reference to FIG. When a musical sound signal received from the outside of the noise removing apparatus 10 is converted into a digital signal by the A / D 130 and a digital signal (hereinafter simply referred to as a digital signal) as the musical sound signal is input to the CPU 100, the CPU 100 The waveform is divided into a plurality of blocks each consisting of a predetermined number of sample values and stored in the RAM 120. Then, CPU 100 obtains the spectrum of the noise waveform of the digital signal and stores the obtained spectrum of the noise waveform in RAM 120 (step S2000).
[0023]
Next, the CPU 100 sets a variable I for designating a target block to which an effect of an acoustic effect is added by processing in the frequency domain among a plurality of blocks obtained by dividing the digital signal to “1” (step S2010). ). The CPU 100 stores the variable I set to “1” in the RAM 120.
[0024]
Next, the CPU 100 executes a process for reading the block specified by the variable I (step S2020). CPU 100 searches RAM 120 and reads variable I. Then, it is determined that the variable I is “1”, and the RAM 120 is searched to read the first block from the front on the time axis among the plurality of blocks obtained by dividing the digital signal.
[0025]
Next, the CPU 100 executes processing for applying a window to the digital signal (step S2030). The CPU 100 reads each block from the RAM 120 and executes a process for setting a window for each block. Here, as an example of the windowing process, as shown in FIG. 4, each of a plurality of blocks obtained by dividing a digital signal is given an overlap corresponding to a half length of the block on the time axis. Executes the process of applying a window.
[0026]
Next, the CPU 100 executes processing for FFT of the digital signal (step S2040). The CPU 100 searches the RAM 120, reads each block, and executes a process of performing FFT on each block.
[0027]
Next, the CPU 100 executes processing for removing the noise waveform from the block designated by the variable I (step S2050). CPU 100 searches RAM 120 and reads the spectrum of the noise waveform detected in step S2000. The noise waveform spectrum is subtracted from the block spectrum read in step S2020. By this process, the waveform of the block from which the noise waveform is removed is acquired.
[0028]
Next, the CPU 100 executes processing in the frequency domain (step S2060). The CPU 100 adds various effects such as acoustic effects by executing processing in the frequency domain on the waveform of the block acquired in step S2050. Examples of the acoustic effect include a low-pass filter, a phaser, a delay reverb, and the like. Then, the CPU 100 stores in the RAM 120 the block that has been processed in this frequency domain.
[0029]
Next, the CPU 100 executes a process of performing inverse FFT on the digital signal (step S2070). The CPU 100 searches the RAM 120, reads each block, and executes a process of performing inverse FFT on each block.
[0030]
Next, the CPU 100 determines whether or not the variable I is 3 or more (step S2080). The CPU 100 searches the RAM 120 to read the variable I, and when the variable I is 3 or more (Yes in step S2080), executes the process in step S2090. If the variable I is not 3 or more (No in step S2080), step S3100 is executed on the assumption that no smoothing process is performed because there is no adjacent block on the front side on the time axis.
[0031]
Next, the CPU 100 executes a smoothing process on the block specified by the variable I (step S2090). The smoothing process will be described in detail later.
[0032]
Next, the CPU 100 executes a process of writing the block waveform as an output waveform (step S3100). The CPU 100 searches the RAM 120, reads the block that has been processed in the frequency domain in step S2060, that is, the block specified by the variable I, and outputs the waveform of this block to the outside of the noise removal apparatus 10 via the D / A 140. Set as output waveform to be sent.
[0033]
Next, the CPU 100 determines whether or not the variable I is a numerical value indicating the final block among a plurality of blocks obtained by dividing the digital signal (step S2110). The CPU 100 searches the RAM 120 and reads the variable I. When the variable I is a numerical value indicating the final block (Yes in step S2110), the CPU 100 reads all the blocks stored in the RAM 120, that is, all the blocks set as output waveforms in step S3100. Then, these blocks are converted into analog signals as musical sound signals by the D / A 140 and transmitted to the outside of the noise removing apparatus 10.
[0034]
When the variable I is not a numerical value indicating the final block (No in step S2110), the CPU 100 changes the variable I, adds “1”, and stores the changed variable I in the RAM 120 (step S2120). And CPU100 repeatedly performs the process after step S2020.
[0035]
Subsequently, the smoothing process in step S2090 described above will be described in detail with reference to the flowchart shown in FIG. 3. First, each of the blocks stored in the RAM 120 performs a windowing process in step S2030. As a result of execution, the blocks are overlapped by a half length on the time axis. For this reason, the first block from the front side on the time axis, the third block separated by two on the first rear side, the second block, the second block, etc. They are adjacent to each other on the time axis.
[0036]
For example, at the connection point when the adjacent first and third blocks are connected, the last amplitude value of the first block does not match the first amplitude value of the third block, which is not valid. There is continuity. Similarly, discontinuities occur at connection points such as between the second and fourth blocks.
[0037]
The smoothing process in step S2090 described above is executed to correct the first and third blocks and the like, and eliminate discontinuities at the connection points between these two blocks. Here, the smoothing process when the variable I is “3” will be described, but the same process is performed even when the variable I is other than “3”.
[0038]
CPU100 performs the process which calculates | requires the amplitude difference of the connection point between the blocks adjacent on a time-axis (step S3000). The CPU 100 searches the RAM 120 to read the block specified by the variable I, that is, the third block from the front on the internal time axis of each block, and the block two blocks before the third block, that is, Read the first block. Then, as shown in FIG. 5, from the amplitude value of the sample at the last position on the waveform of the first block (hereinafter referred to as a sample value), the leading position (first position) on the waveform of the third block. ) Is subtracted to determine the amplitude difference.
[0039]
For example, as shown in FIG. 5, when the sample value at the beginning of the third block is “0” and the sample value at the end of the first block is “−0.3”. The amplitude difference is “−0.3”.
[0040]
Next, the CPU 100 performs a process for obtaining a correction amount in the smoothing process (step S3010). As shown in FIG. 6, the CPU 100 divides the amplitude difference obtained in step S3000 by 2, and stores the numerical value obtained by the division in the RAM 120 as the correction amount corresponding to each of the first and third blocks. . That is, when the amplitude difference is “−0.3”, the correction amount is “−0.15”.
[0041]
Next, the CPU 100 divides the correction amount by the number of smoothing points to obtain a difference value (dBlockSmtDiff) per sample (step S3020). Here, the number of smoothing points is a numerical value indicating the number of samples to be subjected to the smoothing process among the sample values in the block. That is, as shown in FIG. 6, the smoothing score is a numerical value indicating the number of sample values included in the range of the smoothing length indicating the portion of the block on which the smoothing process is performed, and the smoothing score is within the range of the smoothing length. Correction is performed on the number of sample values indicated by.
[0042]
The CPU 100 searches the RAM 120, reads the correction amount, and divides the correction amount by the number of smoothing points to obtain a difference value (dBlockSmtDiff) per sample for the sample to be smoothed. The difference value is a numerical value used for correcting the sample value of one sample. Then, the obtained difference value (dBlockSmtDiff) is stored in the RAM 120. That is, when the correction amount is “−0.15” and the smoothing score is “30”, the difference value per sample (dBlockSmtDiff) is “−0.005”.
[0043]
Next, the CPU 100 sets a variable J for designating a sample to be corrected to “1” (step S3030). The CPU 100 stores the variable J set to “1” in the RAM 120.
[0044]
Next, the CPU 100 executes processing for adding the numerical value calculated based on the difference value and the variable J to the sample value of the sample at the position corresponding to the variable J in the third block (step S3040). The CPU 100 searches the RAM 120 to read the variable J, and uses the variable J to obtain a numerical value (hereinafter referred to as “J ′”) obtained by “smoothing score−J”. Further, the RAM 120 is searched to read a difference value (dBlockSmtDiff), and a numerical value (hereinafter referred to as “J ″”) obtained by “difference value (dBlockSmtDiff) × J” using the difference value (dBlockSmtDiff). Ask. That is, when the smoothing score is “30” and the difference value (dBlockSmtDiff) is “−0.005”, “J ′” is “29” in response to the variable J being “1”. “J ″” becomes “−0.05”.
[0045]
Then, the CPU 100 adds “J ″” to the sample value of the “J ′”-th sample from the top on the time axis of the third block. The CPU 100 corrects the waveform of the third block using the obtained sample value after addition, and stores the corrected third block in the RAM 120. That is, when “J ′” is “29” and “J ″” is “−0.05”, the sample value of the 29th sample from the top on the time axis of the third block is set. “−0.05” is added, and the sample value after this addition is used so that the amplitude value of the 29th sample from the beginning of the third block is reduced by “0.05”. Correct the waveform of the block.
[0046]
Next, the CPU 100 executes a process of subtracting the numerical value calculated based on the difference value and the variable J from the sample value of the sample at the position corresponding to the variable J in the first block (step S3050). The CPU 100 subtracts “J ″” from the sample value of the sample “J ′” before the last on the time axis of the first block. The CPU 100 corrects the waveform of the first block using the obtained sample value after subtraction, and stores the corrected first block in the RAM 120. That is, when “J ′” is “29” and “J ″” is “−0.05”, the sample value of the 29th previous sample from the last on the time axis of the first block “−0.05” is subtracted from the sample value, and the sample value after the subtraction is used so that the amplitude value of the 29th sample from the end of the first block is increased by “0.05”. Correct the waveform of the second block.
[0047]
Next, the CPU 100 changes the variable J, adds “1”, and stores the changed variable J in the RAM 120 (step S3060).
[0048]
Next, the CPU 100 determines whether or not the variable J has the same value as the smoothing score (step S3070). The CPU 100 searches the RAM 120 and reads the variable J. If the variable J has the same value as the smoothing score (Yes in step S3070), all the sample values of the number of samples indicated by the smoothing score are corrected. The process is terminated assuming that the correction of the waveform of the third block has been completed. When all the sample values of the number of samples indicated by the number of smoothing points are corrected, as shown in FIG. 8, the respective waveforms having separated amplitude differences are within the smoothing length range of the first and third blocks. Since the correction is made so as to approach, the amplitude difference is eliminated, and the discontinuity at the connection point between the first and third blocks is eliminated. Further, even if discontinuity occurs between adjacent blocks on the time axis as shown in the waveform before the smoothing process shown in FIG. 9, each waveform is corrected so as to approach each other like the waveform after the smoothing process. Therefore, the amplitude difference disappears and the discontinuity is eliminated.
[0049]
On the other hand, if the variable J is not the same value as the smoothing score (Yes in step S3070), the CPU 100 searches the RAM 120 and searches for the variable J, assuming that all the sample values of the number of samples indicated by the smoothing score are not modified. Reading and using this variable J, the processing after step S3040 is repeated. That is, the CPU 100 sequentially changes the variable J from the 30th to the top side for each of the sample values from the top to the 30th of the third block, and calculates based on the difference value and the variable J. Repeat the process of adding the numerical values. Also for the first block, the variable J is sequentially changed from the 30th to the last side for each of the sample values from the last to the 30th of the first block, and the difference value and the variable The process of subtracting the numerical value calculated based on J is repeated.
[0050]
It should be noted that the variable J is sequentially changed from “1” until the value becomes equal to the number of smoothing points, and the processing in step S3040 and step S3050 is repeated, as shown in FIG. Is equivalent to adding the waveform obtained by sequentially changing the variable J to the waveforms of the first and third blocks.
[0051]
In this way, as a result of applying a window to a certain block of blocks obtained by dividing a digital signal as a musical sound signal into blocks each having a predetermined number of sample values, the block becomes adjacent on the time axis. In the case where both blocks are connected to a block separated by a predetermined number of blocks on the rear side, if discontinuity occurs at the connection point between these two blocks, the CPU 100 performs a smoothing process.
[0052]
Then, as the smoothing process, the CPU 100 executes a process of subtracting and adding all the sample values of the number of samples indicated by the number of smoothing points in the first and third blocks on the time axis. Is used to correct the waveforms of the first and third blocks so that the waveforms with the separated amplitude difference approach each other within the smoothing length range of the first and third blocks. Therefore, the amplitude difference disappears and the discontinuity at the connection point between the first and third blocks is eliminated.
[0053]
In addition, since the sample values of the samples on the waveforms of the first and third blocks are corrected, the processing of performing FFT in the block is performed as in the invention described in JP-T-2002-517021. Without changing the number of samples, it is possible to eliminate the discontinuity without changing the number of samples.
[0054]
Furthermore, unlike the case where processing is performed to form a triangular window or a Hanning window by performing processing for applying a window before performing FFT and performing processing for applying a window after performing inverse FFT, the range indicated by the smoothing length Since the smoothing process is performed only inside, it is possible to eliminate the discontinuity without increasing the amount of calculation.
[0055]
(Other embodiments)
Although the embodiment of the present invention has been described above, various modifications and changes can be made to the above-described embodiment without departing from the gist of the present invention. For example, the smoothing length and the number of smoothing points may be changed as necessary.
[0056]
In the present embodiment, as a process for applying a window, a case has been described in which the window is applied with an overlap corresponding to the length of one half of the block. However, the present invention is not limited to this. When two blocks that are a predetermined number of blocks apart are adjacent to each other even when the window is overlapped by the length of 2 or when the window is opened by another method, etc. In addition, when discontinuity occurs at the connection point between these two blocks, it is possible to apply the noise removal apparatus 10 in the present embodiment.
[0057]
【The invention's effect】
As described above, according to the present invention, when two blocks separated by a predetermined number of blocks that are adjacent on the time axis as a result of applying a window are connected, the connection between those two blocks is connected. When discontinuity occurs at a point, noise that can eliminate the discontinuity without increasing the amount of computation while maintaining the original number of samples by correcting the waveform of both blocks to approach each other A removal device can be provided.
[Brief description of the drawings]
FIG. 1 is a block configuration diagram of a noise removal apparatus 10;
FIG. 2 is a flowchart showing overall processing in the noise removal apparatus 10;
FIG. 3 is a flowchart showing smoothing processing in step S2090.
FIG. 4 is a diagram illustrating a state in which windows are provided with an overlap of a half length of a block as a process of applying a window to each block.
FIG. 5 is a waveform diagram showing a state in which the first and third blocks are adjacent on the time axis.
FIG. 6 is an enlarged view showing the vicinity of a connection point between the first and third blocks in FIG. 5;
FIG. 7 is a waveform diagram showing waveforms for correcting the waveforms of the first and third blocks.
FIG. 8 is a waveform diagram showing a state in which discontinuities are eliminated at connection points between the first and third blocks.
FIG. 9 is a graph showing a state in which discontinuities at connection points between adjacent blocks on the time axis are eliminated.
[Explanation of symbols]
10 Noise removal device
100 CPU
110 ROM
120 RAM
130 A / D (analog / digital converter)
140 D / A (digital / analog converter)

Claims (3)

A block (front block) in a block group obtained by dividing a digital signal generated by sequentially sampling analog signals into blocks each having a predetermined number of sample values, and a result of applying a window to this block and the time axis A noise removing device for performing a smoothing process for eliminating discontinuities at a connection point when both blocks are connected to a block (rear block) that is separated by a predetermined number of blocks on the rear side that comes to be adjacent to each other ,
Correction amount calculation means for obtaining a correction amount when performing the smoothing process at a connection point where discontinuity between the two blocks has occurred;
A difference calculating means for dividing the obtained correction amount by a smoothing score indicating the number of sample values to be subjected to the smoothing processing among the sample values in the block to obtain a difference value per sample;
A process of adding a numerical value calculated based on the difference value and the certain variable to a sample value at a position corresponding to a certain variable in the rear block is defined as an addition process, and from the head of the rear block For each of up to the number of sample values indicated by the number of smoothing points, an addition means for sequentially changing the certain variable from the number of times indicated by the number of smoothing points to the top side and repeatedly performing the addition process When,
A process of subtracting a numerical value calculated based on the difference value and the certain variable from a sample value at a position corresponding to the certain variable in the front block is defined as a subtraction process, and from the end of the front block, For each sample value up to the number before the number indicated by the number of smoothing points, the certain variable is sequentially changed from the number before the number indicated by the number of smoothing points to the end side, and the subtraction process is repeated. And a subtracting means. In the noise removal apparatus of Claim 1,
The correction amount calculating means includes
Means for obtaining an amplitude difference at a connection point where the discontinuity between the two blocks has occurred, and means for obtaining a correction amount when performing the smoothing process by dividing the obtained amplitude difference by two. A noise removing device characterized by the above. In the noise removal apparatus as described in any one of Claim 1 and 2,
The noise removal apparatus according to claim 1, wherein the digital signal is a musical tone signal.

Images