JP2003323194A - System and method for eliminating noise - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a noise elimination system and noise eliminating method for effectively eliminating wideband noise from a narrow band signal such as a stereo sound signal. <P>SOLUTION: When M observation signals x<SB>i</SB>(k) are sequentially inputted to a signal preprocessing part 10 through M channels 2a of inputting part 2 time sequentially, signal preprocessing is conducted so as to eliminate noise from the M observation signals x<SB>i</SB>(k) in a signal preprocessing part 10 having N stage singular value decomposition units 12 cascaded with one another. Then, M signals y<SB>i</SB>(k) subjected to signal preprocessing are inputted to an adaptive signal emphasizing part 20, and the adaptive signal emphasizing part 20 performs adaptive filter processing so as to emphasize the M signals y<SB>i</SB>(k). Thus, a final signal s<SB>i</SB>(k) that has been subjected to noise elimination is outputted. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise removing system and a noise removing method for removing noise from various observed signals, and more particularly to removing wideband noise from narrowband signals such as stereo audio signals. The present invention relates to a noise removal system and a noise removal method that are preferably used in the above.

[0002]

2. Description of the Related Art Communication processing using mobile phones, voice recognition processing, analysis processing of data sent from radar,
Noise removal is indispensable in various signal processing fields such as measurement processing of EEG and electrocardiogram.

Generally, as a method for reducing wideband noise from a narrowband signal such as a voice signal, a non-linear spectral subtraction (NSS) method (see Reference 1) is used. In this method, the spectra corresponding to each of the voice signal component and the noise signal component are individually estimated based on the observed signal in which noise is mixed within a certain time, and the estimated noise spectrum is subtracted from the observed signal. It is a thing.

However, in such a non-linear spectrum subtraction method, when a signal after noise removal is converted from the frequency domain to the time domain, due to an estimation error of the noise spectrum, a secondary instrument such as a kind of musical instrument is used. There is a problem in that noise is generated, and such noise is added instead of reducing wideband noise. Also, SN
When the ratio is low, there is a problem that a part of the voice signal is also removed as the broadband noise is removed. Furthermore, in order to obtain good noise reduction performance, it is necessary to adjust a large number of parameters,
In addition, there is a problem that it is necessary to perform the optimum adjustment for each different environment.

By the way, in recent years, in the field of biomedical engineering, a singular value decomposition method (SV) is applied to a biological signal which is an observed signal.
A method has been proposed in which only specific elements are extracted from a biological signal by applying D: singular value decomposition (see, for example, Reference 2) (see References 3 and 4). These methods use the singular value decomposition method for the observed signal to obtain a signal subspace and a noise subspace.
e) and an orthogonal projection (ONP: orthonormal proj) of the observed signal to the separated signal subspace.
section) to extract a plurality of output signals (noise-free signals) which are time domain signals. Note that such a singular value decomposition method is applied not only to removing noise from a biological signal, but also to removing noise from a voice signal (see References 5 to 10).

As a method using such a singular value decomposition method, there is also proposed a method of extracting a evoked potential of the brain by applying Wiener filtering before applying the singular value decomposition method. (Reference 11).

[0007]

However, in the conventional method using the singular value decomposition method as described above, noise is removed from a plurality of observed signals having correlation with each other such as a stereo audio signal. No consideration is given to such a case, and in such a case, there is a problem that noise cannot be sufficiently removed from a plurality of input observation signals.

Further, the method of performing the Wiener filter processing before applying the singular value decomposition method is also not effective for an observation signal which changes abruptly with time such as a voice signal due to the nature of the Wiener filter. , There is a problem.

The present invention has been made in consideration of the above points, and provides a noise removing system and a noise removing method capable of effectively removing wide band noise from a narrow band signal such as a stereo audio signal. The purpose is to do.

[0010]

As a first solution, the present invention provides an input section having a plurality of channels for inputting a plurality of observation signals, and an input through the plurality of channels of the input section. A signal preprocessing unit that performs signal preprocessing to remove noise from the plurality of observed signals, and an adaptive signal enhancement unit that performs adaptive filter processing to enhance the plurality of signals output from the signal preprocessing unit. And an output unit having a plurality of channels for outputting a plurality of signals output from the adaptive signal enhancing unit.

In the above first solution means,
The signal preprocessing unit separates each of the plurality of observation signals input via the plurality of channels of the input unit into a signal subspace and a noise subspace by a singular value decomposition method, and the signals are separated. It is preferable to have a singular value decomposition unit for extracting a plurality of output signals which are time domain signals by orthogonally projecting the plurality of observation signals onto a signal subspace. Further, the signal preprocessing unit may have a plurality of stages of singular value decomposition units cascaded with each other, which process the plurality of observation signals input via the plurality of channels of the input unit. Well, in this case, the singular value decomposition unit of each stage separates a plurality of input signals corresponding to each of the plurality of channels into a signal subspace and a noise subspace by a singular value decomposition method, and By orthogonally projecting the plurality of input signals onto the generated signal subspace, a plurality of output signals which are time domain signals are extracted.
Here, the signal pre-processing unit shifts at least two output signals belonging to different channels among the plurality of output signals output from the singular value decomposition unit of each stage by a predetermined number of samples from each other by a delay device or a forward device. It is preferable to further have Further, it is preferable that the signal preprocessing unit further includes an amplitude adjusting unit that increases or decreases a plurality of output signals output from a final stage singular value unit of the plurality of stages of singular value decomposition units.

Further, in the above-mentioned first solving means,
It is preferable that the adaptive signal emphasizing unit has an adaptive filter that performs adaptive filter processing on the plurality of signals output from the signal preprocessing unit. In addition, it is preferable that the adaptive signal emphasizing unit further includes a delay unit that delays the plurality of signals output from the signal preprocessing unit and input to the adaptive filter by a predetermined number of samples.

Further, in the above-mentioned first solving means, it is preferable that the adaptive signal emphasizing section is connected in series or in parallel with the signal preprocessing section.

As a second solving means, the present invention includes a step of inputting a plurality of observation signals, a signal preprocessing step of performing signal preprocessing so as to remove noise from the plurality of observation signals, and a signal preprocessing. There is provided a denoising method characterized by including a step of performing adaptive filter processing so as to emphasize a plurality of applied signals and a step of outputting a plurality of signals subjected to the adaptive filter processing.

In the second solving means described above, in the signal preprocessing step, each of the plurality of observed signals is separated into a signal subspace and a noise subspace by the singular value decomposition method, and is also separated. It is preferable to extract a plurality of output signals which are time domain signals by orthogonally projecting the plurality of observation signals onto the signal subspace. Further, in the signal preprocessing step, noise may be removed from the plurality of observed signals by applying a singular value decomposition method to the plurality of observed signals a plurality of times. The singular value decomposition method of each time separates a plurality of input signals corresponding to the plurality of observed signals into a signal subspace and a noise subspace by the singular value decomposition method, and the plurality of input signals corresponding to the separated signal subspaces. By orthogonally projecting the input signal of
A plurality of output signals which are time domain signals are extracted.

According to the present invention, the signal preprocessing unit performs the signal preprocessing so as to mainly remove noise from the plurality of observed signals, and then the adaptive signal emphasizing unit outputs the plurality of signals output from the signal preprocessing unit. Since the adaptive filter processing is applied to emphasize the signal, the waveform of the input observation signal is good even when removing noise from multiple observation signals that have a correlation with each other, such as a stereo audio signal. It is possible to effectively remove only the noise while keeping it at.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

First, referring to FIG. 1, the overall configuration of an embodiment of a noise elimination system according to the present invention will be described.

As shown in FIG. 1, the noise elimination system 1 according to the present embodiment has M observation signals x _i (k) (i = 1, 2, ..., M; k) detected by a sensor or the like. Removes noise from the input section 2 having M channels 2a for inputting discrete time) and the M observation signals x _i (k) inputted via the M channels 2a of the input section 2. Signal pre-processing unit 10 for performing signal pre-processing so that M signals y _i (k) (i = 1, 1) output from the signal pre-processing unit 10
2, ..., M; k is a discrete time) and an adaptive signal enhancement unit (adaptive signal enha) that performs adaptive filter processing
ncer) 20 and M channels 3a for outputting M signals s _i (k) (i = 1, 2, ..., M; k is a discrete time) output from the adaptive signal emphasizing unit 20. And an output unit 3 having the same.

(Signal Preprocessing Unit) FIG. 2 is a diagram showing a detailed configuration of the signal preprocessing unit 10 shown in FIG. As shown in FIG. 2, the signal processing unit 10 includes a noise removing unit 12 that removes noise from M observation signals x _i (k) input through the M channels 2 a of the input unit 2, and a noise removing unit 12. The amplitude adjusting unit 14 increases or decreases the amplitude of the M signals y _i (k) output from the removing unit 12.

Of these, the noise elimination unit 12 processes M observation signals x _i (k) input via the M channels 2a of the input unit 2, and is connected in cascade to each other. It has a singular value decomposition unit (SVD unit) 12a. The singular value decomposition unit 12a at each stage separates the M input signals corresponding to each of the M channels 2a into a signal subspace and a noise subspace by the singular value decomposition method, and the separated signals. By orthogonally projecting M input signals to the subspace, M output signals that are time domain signals are extracted. The basic method of the singular value decomposition method performed by the singular value decomposition unit 12a at each stage is generally the same as that of the existing one (see, for example, Documents 11 and 12).

The details of the singular value decomposition method performed by the singular value decomposition unit 12a at each stage will be described below.

First, input signal data is converted into an L × M matrix.
The matrix X = [x ₁，x _Two，… ，x _M]]. Na
Oh, the column vector of matrix Xx _i(I = 1, 2, ..., M)
Isx _i= [X_1i, X_2i, ..., x_Li]^T  (T is
It represents a transposed vector). In addition, in this specification
The underlined letter represents a vector.

Such a matrix X is represented by the following equation (1) by the singular value decomposition method, where M <L.

[0025]

[Equation 1]

Where the matrices U and V are U =
[ U ₁ , u ₂ , ..., u _M ] ε ^{RL × M} , V = [ v ₁ , v
₂ , ..., V _M ] εR ^{M × M} , and U ^T U =
Let I _M and V ^T V = I _M be orthogonal matrices. Also, the matrix Σ is Σ = diag (σ ₁ , σ ₂ ,
, Σ _M ) εR ^{M × M} , and σ ₁ ≧ σ ₂ ≧ ... ≧ σ _M ≧
It shall be 0. The column vectors included in the matrices U and V are called the left singular value vector and the right singular value vector of X, respectively. The diagonal components of the matrix Σ are called singular values and contain information about the number of signals, energy, noise level, and the like.

When the SN ratio is sufficiently high, the matrix X is decomposed as in the following equation (2).

[0028]

[Equation 2]

Here, the matrix Σ _s represents the maximum singular values associated with s signal sources, and the matrix Σ _n represents the (M−s) singular values associated with noise. Also, the matrices U _s and V _s both have s singular value vectors associated with the source, and the matrices U _n and V _n both have (M−s) singular value vectors associated with the noise. ing. The subspace spanned by the column vectors of the matrix U _s is called the signal subspace, and the subspace spanned by the column vectors of the matrix U _n is called the noise subspace.

Since the signal subspace and the noise subspace are theoretically orthogonal to each other, the principle of least-squares approximation is obtained by orthogonally projecting data with noise, which is an observation signal, onto the signal subspace. Can be used to remove noise.

That is, the output after the noise is removed
The signal data is converted into the matrix Y = [y ₁，y _Two，… ，y _M] (Column
vectory _i(I = 1, 2, ..., M) isy _i=
[Y_1i, Y _2i, ..., y_Li]^T )
Then, by the orthogonal projection described above, the matrix Y is given by the following equation (3).
Given more.

[0032]

[Equation 3]

The above equation (3) is simply represented by the following equation (4) by the characteristic (orthogonal characteristic) of the vector representing the signal subspace.

[0034]

[Equation 4]

That is, in the singular value decomposition unit 12a at each stage shown in FIG. 1, L samples of M input signals input from M channels 2a are set as one frame, and the L × M input signals are input. The singular value decomposition method is applied according to the above equation (4) to the matrix X including.

Here, in the existing method using the conventional frame arithmetic expression (see, for example, Documents 11 and 12), FIG.
As shown in (b), the matrix Y corresponding to the matrix X is obtained for each L sample frame unit that does not overlap each other, and L × M output signals are extracted. As shown in (a), the matrix Y corresponding to the matrix X may be obtained in frame units of L samples that overlap each other in a manner that they are offset by one sample.

In the method shown in FIG. 3A, only some elements (see reference numeral 31 in FIG. 3B) of the matrix Y as the output result by the singular value decomposition method are the final output signals. The remaining elements are all updated every time the discrete time k increases by 1 (time advances by one sample). On the other hand, in the method shown in FIG. 3B, all the elements (see reference numeral 32 in FIG. 3B) of the matrix Y as the output result by the singular value decomposition method are used as the final output signal. According to the method shown in FIG. 3A, in the singular value decomposition unit 12a at each stage, the matrix U _s representing the signal subspace is updated at each point of the discrete time k, and the input signal On the other hand, it is possible to function in the same way as a filter. Therefore, the method shown in FIG. 3 (a) is particularly preferably used in a configuration in which a plurality of stages of singular value decomposition units 12a of the signal preprocessing unit 12 as shown in FIG. 2 are cascade-connected.

The amplitude adjusting section 14 has M amplitude adjusters 14a corresponding to the respective channels 3a of the output section 3. In addition, the code c _i attached to each amplitude adjuster 14a
(I = 1, 2, ..., M) represents a coefficient by which each of the M signals output from the noise removing unit 12 is multiplied, and the value of the coefficient is provided in the subsequent stage of the amplitude adjusting unit 14. The adaptive signal enhancing unit 20 (see FIG. 1) connected is appropriately adjusted so that the signal is favorably enhanced.

(Adaptive Signal Enhancement Unit) As shown in FIG. 1, an adaptive signal enhancement unit 20 is connected in series after the signal preprocessing unit 10.

Here, the adaptive signal emphasizing unit 20 performs adaptive filter processing on each of the M signals y _i (k) output from the signal preprocessing unit 10 for the respective signals y _i (k). It has an adaptive filter 20a to be applied and a delay device 20b that delays the signal y _i (k) input to the adaptive filter 20a by a predetermined number of samples l ₀ .

Of these, the adaptive filter 20a is a predetermined filter.
Number of samples l₀Signal y delayed only _iSuitable for (k)
Signal s after final noise removal after adaptive filtering
_iIt is for outputting (k) and has an internal odor
Error output e_i(K) (observation signal x_i(K) and noise
Signal s after removal_i(Difference from (k))
Algorithm (eg NLMS (normalized LMS) method)
The coefficients used in the adaptive filter are updated according to
It Note that the adaptive filter processing itself is existing (for example,
For example, refer to Reference 13).

Further, the delay unit 20b includes the signal preprocessing unit 10
The signal y _i (k) output from the input signal and input to the adaptive filter 20a is intentionally delayed by l ₀ (= (L _f −1) / 2) samples (L _f is the length of the adaptive filter). . In this way, by delaying the signal y _i (k) input to the adaptive filter 20a by the predetermined number of samples l ₀ , the adaptive filter processing performed by the adaptive filter 20a effectively reduces the signal y _i (k). In addition to removing noise, the amplitude and phase shift between channels, which is one of the characteristics of stereo audio signals, is effectively restored.

Next, the operation of this embodiment having such a configuration will be described.

In the noise reduction system 1 shown in FIG.
The M observation signals x _i (k) detected by the sensor are
The signals are sequentially input to the signal preprocessing unit 10 in time series through the M channels 2a of the input unit 2.

The signal preprocessing unit 10 performs signal preprocessing so as to remove noise from the M observed signals x _i (k).

Specifically, in the signal preprocessing section 10,
As shown in FIG. 2, the M observation signals x _i (k) input through the M channels 2a of the input unit 2 are input to the singular value decomposition unit 12a of the first stage. In the singular value decomposition unit 12a in the first stage, the M input signals corresponding to each of the M channels 2a are divided into a signal subspace and a noise subspace by the singular value decomposition method. By separating and orthogonally projecting the M input signals to the separated signal subspace, M output signals which are time domain signals are extracted, and the extracted M output signals are extracted. It is sent to the singular value decomposition unit 12a at the next stage (second stage).

Thereafter, similarly, in the second to Nth singular value decomposition units 12a, the M output signals output from the preceding singular value decomposition unit 12a are used as input signals, and the first stage singular value decomposition is performed. Processing similar to that of the value decomposition unit 12a is performed.

As a result, the N-th stage singular value decomposition unit 12a finally outputs a plurality of signals from which noise has been removed from the M observed signals x _i (k).

Here, the plurality of signals thus output are input to the respective amplitude adjusters 14a of the amplitude adjusting section 14, and the amplitudes of the respective signals are multiplied by the coefficient c _i .

Thereafter, each amplitude adjuster 1 of the amplitude adjusting unit 14
The signals whose amplitudes have been increased or decreased by 4a are output from the signal preprocessing unit 10 as M signals y _i (k) that have been subjected to signal preprocessing, and are output to the adaptive signal enhancement unit 20 as shown in FIG. Is entered.

In the adaptive signal emphasizing section 20, adaptive filter processing is performed so as to emphasize the M signals y _i (k) which have been subjected to the signal preprocessing.

Specifically, in the adaptive signal emphasizing unit 20, the M signals y output from the signal preprocessing unit 10 are output.
After delaying each _i (k) by a predetermined number of samples l ₀ , the respective signals y _i (k) are input to the adaptive filter 20a, and in this adaptive filter 20a, a signal y _i delayed by a predetermined number of samples l _0. The signal s after the final noise removal is obtained by applying adaptive filtering to (k).
_i (k) is output. In the adaptive filter 20a, the error output e _i (k) (observation signal x
Based on the difference between _i (k) and the noise-removed signal s _i (k)), the coefficient used in the adaptive filter is updated according to a predetermined algorithm (for example, the NLMS method).

As described above, according to the present embodiment, the signal preprocessing unit 10 performs signal preprocessing so as to mainly remove noise from the plurality of observed signals x _i (k), and then the adaptive signal enhancement unit 20. Accordingly, since the adaptive filter processing is performed so as to emphasize the plurality of signals y _i (k) output from the signal preprocessing unit 10, a plurality of observed signals x such as a stereo audio signal having a correlation with each other are obtained. Even when noise is removed from _i (k), the input observed signal x
Only noise can be effectively removed while the waveform of _i (k) is kept good.

Further, according to the present embodiment, in the signal pre-processing section 10, a plurality of observation signals x _i are made by the singular value decomposition units 12a of a plurality of stages which are cascade-connected to each other.
Since the singular value decomposition method is applied to (k) a plurality of times, even when noise is removed from a plurality of observed signals x _i (k) that are correlated with each other, such as a stereo audio signal. , It is possible to effectively remove only noise while maintaining a good waveform of the input observed signal x _i (k). In addition, a plurality of singular value decomposition units 12a cascade-connected to each other allows a plurality of observation signals x
_{Since the} singular value decomposition method is applied to _i (k) a plurality of times, the number of times of the singular value decomposition method applied to a plurality of observation signals x _i (k) can be set arbitrarily. Even if the number of sensors and the like is small and the number (M) of observation signals x _i (k) is small, the singular value decomposition unit 12a
The noise reduction performance can be easily improved by increasing the number of stages (N). Furthermore, since the singular value decomposition method is used as a noise removal method applied to a plurality of observed signals x _i (k), even if the positions and numbers of signal sources and noise sources are not known in advance, The signal component and the noise component can be easily separated, and the noise can be easily removed from the input observation signal x _i (k) regardless of the position and number of microphones (sensors) for detecting the voice signal. can do. That is, the existing speech separation method (see References 14 to 18).
For example, it is effective when multiple microphones (sensors) are installed near the signal source, but if all microphones are installed near the signal source (for example, the speaker) or noise sources There is a problem that it does not work well when the microphones are moving away, and there is also a problem that it does not work normally when the number of signal sources exceeds the number of microphones (sensors). Therefore, such a problem does not occur. In addition, even for an observation signal x _i (k) whose amplitude changes rapidly with the passage of time, such as a voice signal, without using a mechanism such as a voice detector (voice activity detector). , The noise can be easily removed from the observed signal x _i (k).

In the above-described embodiment, the singular value decomposition units 12a in each stage of the noise elimination unit 12 of the signal preprocessing unit 10 are directly connected, but as shown in FIG. Delayers 15 and 16 or advancers 17 and 18 may be inserted between the singular value decomposition units 12a of each stage of the noise removing unit 12.

The signal pre-processing unit 10 'shown in FIG. 4 is the same as the signal pre-processing unit 10 shown in FIG. 17 and 18 are inserted. The signal preprocessing unit 10 'shown in FIG. 4 removes noise from the strongly correlated 2-channel observation signals x ₁ (k) and x ₂ (k) such as a stereo audio signal. However, the basic configuration is substantially the same as that of the signal preprocessing unit 10 shown in FIG. In the signal preprocessing unit 10 'shown in FIG. 4, the same parts as those of the signal preprocessing unit 10 shown in FIG. 2 are designated by the same reference numerals, and detailed description thereof will be omitted.

As shown in FIG. 4, the noise eliminator 12 delays the two-channel signals output from the singular value decomposition unit 12a at each stage by p samples from each other.
16 and advancers 17, 18. Specifically, as shown in FIG. 4, the first channel (the channel corresponding to the observation signal x ₁ (k) is provided at the subsequent stage of each of the singular value decomposition units 12a of the first to (N / 4) th stages. ), A delay device 15 is provided on the upper side of the (N / 4 + 1) th stage to the (N / th stage).
A delay device 16 is provided on the second channel (the channel corresponding to the observation signal x ₂ (k)) after the singular value decomposition unit 12a of the 2) th stage. Further, a forwarder 17 is provided on the first channel after the singular value decomposition unit 12a of each of the (N / 2 + 1) th stage to the (3N / 4) th stage and the (3N / 4 + 1) th stage to the 3rd stage. A forwarder 18 is provided on the second channel after the N singular value decomposition units 12a.

As described above, the signal preprocessing unit 1 shown in FIG.
According to 0 ', the signal of one channel is compared with the signal of the other channel by the delay units 15 and 16 or the forwarder 1.
Since p samples are shifted by 7 and 18, only uncorrelated signal components can be effectively weakened. That is, in a two-channel observation signal with strong correlation such as a stereo audio signal, the audio signal components have a strong correlation with each other on the time axis, while the noise signal components to be removed include white noise. Since there is often no correlation, the noise removal system 10 'as shown in FIG. 4 can weaken only the correlation of the noise signal component which is white noise while maintaining the correlation of the voice signal component to some extent. It is possible to more efficiently remove noise from the observed signal by the singular value decomposition method. At this time, if the sampling frequency (sampling rate / interval) is sufficiently high, the estimation error of the signal subspace by the singular value decomposition method does not pose a problem so much. 2-channel signals y ₁ (k) and y ₂ (k) output to
However, the waveforms of the input observation signals x ₁ (k) and x ₂ (k) can be kept good.

In the signal preprocessing unit 10 'shown in FIG. 4, the mode in which the delay units 15 and 16 and the forward units 17 and 18 are inserted between the singular value units 12a in each stage is arbitrary, but the final mode is not specified. Output signals y ₁ (k) and y ₂ (k)
It is preferable to insert the same number of delay devices and forward devices in each channel in consideration of the consistency of time.

Further, in the above-described embodiment, as a signal preprocessing method in the signal preprocessing section 10, a plurality of stages of singular value decomposition units 1 cascade-connected to each other are used.
2a uses a method of applying the singular value decomposition method to a plurality of observed signals x _i (k) a plurality of times, but the method is not limited to this, and a single singular value decomposition unit can be used to generate a plurality of observed signals x i It is also possible to use a method of applying the singular value decomposition method to _i (k), and other than that, a nonlinear spectrum subtraction method (Reference 1) or a MUSIC method (Reference 1)
It is possible to use various methods such as 9).

Further, in the above-mentioned embodiment,
Although the adaptive signal emphasizing unit 20 is connected in series to the signal preprocessing unit 10, the invention is not limited to this, and the signal preprocessing unit 10 may be connected in parallel to the signal preprocessing unit 10 as in the noise removal system 1 ′ shown in FIG. It may be connected. The basic configuration of each part of the noise removal system 1'shown in FIG. 5 is substantially the same as that of the noise removal system 1 shown in FIG. However, in the case of the noise removal system 1 ′ shown in FIG. 5, a plurality of amplitude adjusters 21 that increase or decrease the amplitude of the signal input to the adaptive signal emphasizing unit 20.
An amplitude adjusting unit 21 having a is provided.

[0062]

EXAMPLES Next, specific examples of the above-described embodiment will be described.

An experiment for removing noise from a two-channel stereo audio signal was conducted using the noise removing system as shown in FIG. In addition, here, as the stereo audio signal, 3 speakers (2 male and 1 female) were recorded (sampled at 48 kHz) in an anechoic room.
Three types of stereo audio signals were prepared, and the following three types of noise (noise for two channels) were added to each of these three types of stereo audio signals.

(1) Uncorrelated noise (2) A fixed filter obtained by assuming an impulse response of an anechoic chamber from a single white noise source provides a certain degree of cross-correlation.
n) with noise (3) Periodic and broadband noise modeling car engine sound

Then, an experiment for removing noise from the above-mentioned nine types of stereophonic audio signals with noise was conducted using the noise removing system having the configuration shown in FIG. Here, in the signal preprocessing unit, the number of stages N of the singular value decomposition unit is 4, the length L of the analysis matrix used in each stage of the singular value decomposition unit is 32, and the value of the coefficient c _i is 0.1. And As for the adaptive signal emphasizing unit, the length of the adaptive filter is 51, and the algorithm for updating the coefficient of the adaptive filter is the NLMS method.

As a result, noise was sufficiently removed from any of the stereo audio signals with noise, and a signal close to the original audio waveform was obtained.

As a comparative experiment, a delay device and a forward device (the number of samples p = 1) are provided between the singular value decomposition units at each stage.
Or, by inserting 6) and performing an experiment to remove noise from the above-mentioned nine types of stereophonic audio signals with noise, for any stereophonic audio signal with noise,
The noise reduction performance was improved when the delay device and the forward device were inserted, compared to when the delay device and the forward device were not inserted.

As a comparative experiment, an experiment for removing noise from the above-mentioned nine types of stereo audio signals with noise was performed by using the conventional non-linear spectrum subtraction method (NSS). Regarding the voice signal, the voice waveform is also greatly deformed, and in addition, a kind of secondary noise such as a musical instrument is added.

The experimental results will be described in detail below, taking one stereo audio signal of the above-described nine types of stereo audio signals with noise as an example.

FIG. 6 shows one stereo audio signal (a stereo audio signal recorded by one male plus uncorrelated noise) among the above-mentioned nine types of stereo audio signals with noise. It is a figure which shows the experimental result which was.
6 (a) is a waveform of a stereo audio signal before noise is added, FIG. 6 (b) is a waveform of a noise signal added to the stereo audio signal, and FIG. 6 (c) is a nonlinear spectrum subtraction method (N
Waveform of the stereo audio signal after the noise is removed by SS), FIG. 6D shows the waveform of the stereo audio signal after the noise is removed only by the signal preprocessing unit (single-valued singular value decomposition unit), 6 (e) is a waveform of a stereo audio signal after noise is removed only by the signal preprocessing unit (four-step singular value decomposition unit), and FIG. 6 (f) is a signal preprocessing unit (four-step singular value decomposition unit). 3 is a waveform of a stereo audio signal after noise is removed by a combination of a decomposing unit) and an adaptive signal emphasizing unit.

As can be seen by comparing FIGS. 6 (d) (e) (f), the speech waveform shown in FIG. 6 (e) is shown in FIG. 6 (a) rather than that shown in FIG. 6 (d). 6A is closer to the original speech waveform shown in FIG. 6F, and the speech waveform shown in FIG. 6F is closer to the original speech waveform shown in FIG. 6A than the speech waveform shown in FIG. 6E. , It was also in agreement with the result of actually viewing the signal after noise removal. In the case shown in FIG. 6 (e) (when only the signal preprocessing unit is used), the stereo sound image is erased instead of the noise being removed, and it sounds like two-channel monaural sound. In the case shown in FIG. 6 (f) (when the adaptive signal emphasizing unit is combined), noise removal is improved at a level of 2 to 3 dB, and the stereo sound image is also restored. It should be noted that such improvement is due to gain (segmental ga
in) and cepstral distance
It was also confirmed by the experimental results by one objective measurement method (see, for example, References 20 and 21).

The speech waveform shown in FIG. 6 (c) is as shown in FIG. 6 (e).
Compared with the speech waveform shown in (f), it looks similar to the original speech waveform shown in FIG. 6 (a), but when the signal after noise removal is actually viewed, as described above, the There was noise like some kind of musical instrument.

[0073]

[Reference] [1] R. Martin, `` Spectral subtraction
based on minimum statistics, '' Proc. EUSIPCO-94, p
p. 1182-1185, Edinburgh, 1994. [2] GH Golub and CF Van Loan, Matrix Computa
tion, 3rd Ed., The Johns Hopkins Univ. Press, Balt
imore and London, 1996. [3] PA Karjalainen, JP Kaipio, AS Koistin
en and M. Vuhkonen, `` Subspace regularization metho
d for the single-trial estimation of evoked potent
ials, '' IEEE Trans. Biomed. Eng., Vol. 40, pp. 849
-860, July 1999. [4] T. Kobayashi and S. Kuriki, `` Principle compon
ent elimination method for the improvement of S / N
in evoked neuromagnetic field measurements, '' IEEE
Trans. Biomed. Eng., Vol. 46, pp. 951-958, Aug. 1
999. [5] F. Asano, S, Hayamizu, T. Yamada, and S. Nakam
ura, `` Speech enhancement based on the subspace me
thod, '' IEEE Trans. Speech, Audio Proc., Vol. 8, N
o. 5, pp. 497-507, Sept. 2000. [6] M. Dendrinos, S. Bakamidis, and G. Carayannis,
`` Speech enhancement from noise: a regenerative a
pproach, '' Speech Communication, Vol. 10, pp. 45-5
7, Feb. 1991. [7] S. Doclo and M. Moonen, `` SVD-based optimal fi
ltering with applications to noise reduction in sp
eech signals, '' IEEE Workshop on App., Sig., Pro
c., to Audio, Acoust., pp. 143-146, New Paltz, NY,
USA, Oct. 1999, also in internal report, KU Leu
ven, Apr. 1999. [8] Y. Ephraim and HLV Trees, `` A signal subs
pace approach for speech enhancement, '' IEEE Tran
s. Speech, Audio Proc., Vol. 3, No. 4, pp.251-266,
July 1995. [9] PSK Hansen, `` Signal subspace methods for
speech enhancement, '' Ph.D. Thesis, Technical Uni
v. of Denmark, Lyngby, Denmark, Sept. 1997. [10] SH Jensen, PC Hansen, SD Hansen, and
JA Sorensen, `` Reduction of broad-band noise i
n speech by truncated QSVD, '' IEEE Trans.Speech, A
udio Proc., Vol. 3, pp. 439-448, Nov. 1995. [11] A. Cichocki, RR Gharieb, and T. Hoya, `` Ef
ficient extraction of evoked potentials by combina
tion of Wiener filtering and subspace methods, '' i
n Proc. ICASSP-2001, Salt Lake City, May 2001. [12] PK Sadasivan and DN Dutt, `` SVD based t
echnique for noise reduction in electroencephalogr
aphic signals, '' Signal Processing, Vol. 55, No.2,
pp. 179-189, 1996. [13] S. Haykin, `` Adaptive Filter Theory, '' 2nd E
d., Englewood Cliffs, NJ: Prentice-Hall, 1991. [14] S. Haykin, `` Unsupervised adaptive filterin
g, '' Volume I & II, John Wiley & Sons, Inc, 2000. [15] S. Amari and A. Cichocki, `` Adaptive blind si
gnal processing-neural network approaches '', Pr
oc. IEEE, Vol. 86, No. 10, pp. 2026-2048, Oct. 199
8. [16] K. Torkkola, `` Blind separation of delayed so
urces based on information maximization, '' Proc.IC
ASSP-96, pp. 3509-3512, 1996. [17] HL Nguyen Thi and C. Jutten, `` Blind sourc
e separation for convolved mixtures, '' Signal Proc
essing, Vol. 45, No. 2, pp. 209-229, 1995. [18] C. Jutten and J. Herault, `` Blind separation
of sources, part I: an adaptive algorithm based on
neuromimetic architecture, '' Signal Processing, Vo
l. 24, No. 1, pp. 1-10, 1991. [19] T. Murakami, M. Namba, T. Hoya, and Y. Ishid
a, `` Speech Enhancement Using MUSIC (MUltiple SIgn
al Classification) Algorithm, '' Proc. IASTED 2001,
pp. 213-216, Rhodes, Greece, July 2001. [20] JR Deller, Jr., JG Proakis, and JH
L. Hansen, `` Discrete-time processing of speech si
gnals, '' Macmillan Publishing Company, 1993. [21] R. Le Bouquin-Jennes, A. Akbari Azirani, and
G. Faucon, `` Enhancement of Speech Degraded by Coh
erent and Incoherent Noise Using a Cross-Spectral
Estimator '', IEEE Trans. On Speech, Audio Proc., V
ol. 5, No. 5, pp. 484-487, Sept. 1997.

[0074]

As described above, according to the present invention, wide band noise can be effectively removed from a narrow band signal such as a stereo audio signal.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a noise reduction system according to the present invention.

FIG. 2 is a block diagram showing details of a signal preprocessing unit of the noise reduction system shown in FIG.

FIG. 3 is a diagram for explaining an outline of a singular value decomposition method used in the signal preprocessing unit shown in FIG.

FIG. 4 is a block diagram showing a modification of the signal preprocessing unit shown in FIG.

FIG. 5 is a block diagram showing another embodiment of the noise reduction system according to the present invention.

FIG. 6 is a diagram showing an experimental result using the noise reduction system according to the specific example of the present invention.

[Explanation of symbols]

1,1 'noise removal system 2 Input section 2a channel 3 Output section 3a channel 10, 10 'signal preprocessing unit 12 Noise removal section 12a Singular value decomposition unit 14 Amplitude adjuster 14a Amplitude adjuster 15, 16 delay device 17,18 Advancer 20 Adaptive signal enhancement section 20a Adaptive filter 20b delay device 21 Amplitude adjuster 21a Amplitude adjuster

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Takahiro Murakami             871-174 Higashifukai, Nagareyama City, Chiba Prefecture (72) Inventor Yoshihisa Ishida             1-19-1 Kuriki, Aso-ku, Kawasaki-shi, Kanagawa F-term (reference) 5D015 CC02 CC03 EE05

Claims (12)

[Claims] 1. An input unit having a plurality of channels for inputting a plurality of observation signals, and a signal pre-signal for removing noise from the plurality of observation signals input via the plurality of channels of the input unit. A signal pre-processing unit that performs processing, an adaptive signal enhancement unit that performs adaptive filter processing to enhance the plurality of signals output from the signal pre-processing unit, and a plurality of signals output from the adaptive signal enhancement unit And a output section having a plurality of channels for performing noise reduction. 2. The signal preprocessing unit separates each of the plurality of observation signals input via the plurality of channels of the input unit into a signal subspace and a noise subspace by a singular value decomposition method. Along with, a singular value decomposition unit for extracting a plurality of output signals which are signals in the time domain by orthogonally projecting the plurality of observed signals onto the separated signal subspace is provided. 1
The noise reduction system described in. 3. The signal preprocessing unit includes a plurality of stages of singular value decomposition units cascaded with each other, for processing the plurality of observation signals input via the plurality of channels of the input unit. , The singular value decomposition unit of each stage separates a plurality of input signals corresponding to each of the plurality of channels into a signal subspace and a noise subspace by a singular value decomposition method, and The noise removal system according to claim 1, wherein a plurality of output signals which are time domain signals are extracted by orthogonally projecting the plurality of input signals. 4. The delay unit for shifting at least two output signals belonging to different channels among the plurality of output signals output from the singular value decomposition unit of each stage by a predetermined number of samples from each other, The denoising system according to claim 3, further comprising a forwarder. 5. The signal preprocessing section further comprises an amplitude adjusting section for increasing / decreasing a plurality of output signals output from a final stage singular value unit of the plurality of stages of singular value decomposition units. The noise reduction system according to claim 3 or 4. 6. The adaptive signal emphasizing section has an adaptive filter for performing adaptive filter processing on a plurality of signals output from the signal pre-processing section. The noise reduction system described in. 7. The adaptive signal emphasizing unit further comprises a delay device for delaying the plurality of signals output from the signal preprocessing unit and input to the adaptive filter by a predetermined number of samples. Item 7. The noise reduction system according to Item 6. 8. The noise elimination system according to claim 1, wherein the adaptive signal enhancing unit is connected in series to the signal preprocessing unit. 9. The noise removing system according to claim 1, wherein the adaptive signal enhancing unit is connected in parallel to the signal preprocessing unit. 10. A step of inputting a plurality of observation signals, a signal preprocessing step of performing signal preprocessing so as to remove noise from the plurality of observation signals, and a step of emphasizing the plurality of signals subjected to the signal preprocessing. And a step of outputting a plurality of signals that have been subjected to the adaptive filter processing. 11. In the signal preprocessing step, each of the plurality of observed signals is separated into a signal subspace and a noise subspace by a singular value decomposition method, and the plurality of signal subspaces are separated from each other. The noise removal method according to claim 10, wherein a plurality of output signals, which are signals in the time domain, are extracted by orthogonally projecting the observation signal. 12. In the signal preprocessing step, noise is removed from the plurality of observed signals by applying the singular value decomposition method to the plurality of observed signals a plurality of times. , A plurality of input signals corresponding to the plurality of observed signals are separated into a signal subspace and a noise subspace by a singular value decomposition method, and the plurality of input signals are orthogonally projected to the separated signal subspace. 11. The noise removing method according to claim 10, wherein a plurality of output signals, which are time domain signals, are extracted thereby.