JP3040651B2 - Signal processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a signal processing apparatus having a function of removing a signal component (noise) other than a signal of a target factor from a signal including a plurality of factors.

[0002]

2. Description of the Related Art Conventionally, in an abnormality detection device, a monitoring device, an inspection device, and the like using a measurement signal such as a monitor signal or a sensor signal, a signal processing device for removing noise included in the measurement signal in order to improve accuracy. Is used.

A conventional signal processing apparatus for removing noise contained in a measurement signal as described above uses a teacher signal as shown in FIG. FIG. 5A shows a process for generating a filter by learning, and FIG. 5B shows a process for filtering using the filter generated in the process of FIG.

First, as shown in FIG. 5A, a teacher signal Sj (t) of a target signal and a teacher signal Ni (t) of a noise are supplied to a filter device 11, and the original signal is processed by the filter device 11. Generate the filter 12.

At the time of actual filtering, a mixed signal V (t) in which a target signal and noise are mixed is input to the filter device 1 as shown in FIG. V ′ (t) is obtained.

[0006] As described above, the conventional signal processing apparatus mainly uses a time-series signal as a filter operation target as it is. In addition, there is another conventional signal processing apparatus in which a signal is once subjected to fast Fourier transform, filtered, and then inversely transformed to obtain a desired signal. FIG. 6 shows an example of a bandpass filter. That is, the mixed signal V (t) is converted to FFT (Fast F
ourier Tranform: Fast Fourier Transform) Input to the device 21 to obtain Fourier coefficients. Then, the Fourier coefficients are input to the filtering device 22 and subjected to filtering to remove noise, and then subjected to inverse transform by the inverse FFT device 23 to obtain a desired output signal V ′ (t).

[0007]

The conventional signal processing apparatus shown in FIG. 5 separates a target signal and noise only by a time-axis waveform (original signal waveform) as shown in FIG. 7A. Also, the signal processing device shown in FIG. 6 separates only the frequency axis as shown in FIG. 7B, and it is difficult to obtain a high separation capability in any case.

When an attempt is made to separate and extract only a target signal from a measurement signal in which an extraction target and noise are mixed, for example,
(a) the waveform of the target signal is deformed by the noise signal,
(b) a small S / N (target signal amplitude / noise amplitude) ratio;
In such a case, since a high separation capability is required, there is a problem that it is difficult to separate a target signal with the conventional signal processing device.

The signal processing apparatus shown in FIG. 6 removes noise by using a technique more advanced than the apparatus shown in FIG. 5, that is, filtering the mixed signal V (t) by performing a fast Fourier transform and then filtering. The result is a method that does not have a time axis. Therefore, the time transition of the conversion result cannot be learned, and a high separation ability cannot be obtained.

Conventionally, a wavelet transformation is known as a signal transformation technique, and there is an example in which the technique is applied to signal discrimination. However, this technique compares wavelet transform coefficients to discriminate the type of signal. Therefore, application to a filtering device has not been considered yet.

The present invention has been made in view of the above-mentioned circumstances. By applying a wavelet transform / inverse transform as a pre-processing / post-processing method of filtering, noise and a target signal overlap to deform the signal shape. It is an object of the present invention to provide a signal processing device capable of reliably removing noise even in a case where the S / N ratio of a measurement signal is small.

[0012]

A signal processing apparatus according to the present invention comprises a noise teacher signal Ni (t) (i is an arbitrary number of 1 or more) and a target signal teacher signal Sj (t) (j is 1). When the above arbitrary number is given, the wavelet transform coefficient sequence WS + N (t, a) (where a:
A wavelet transform means for obtaining a wavelet transform coefficient sequence WS (t, a) of only the scale signal, t: time) and the teacher signal of the target signal; and a wavelet transform coefficient sequence WS + N (t, a) output from the wavelet transform means. a)
Given a filter generation means for generating a wavelet coefficient filter F which is closest to the above-mentioned wavelet transform coefficient sequence WS (t, a), and a time-series mixed signal V (t) in which noise and a target signal are mixed, Means for obtaining a wavelet transform coefficient sequence W V (t, a) of the mixed signal V (t) by the wavelet transform means;
Filtering means for filtering V (t, a) by the filter F obtained by the filter generating means to generate a coefficient sequence W ^f V (t, a); and a coefficient sequence W ^f V ( t, a) is subjected to inverse wavelet transform to remove noise V ^f
(T) output wavelet inverse transform means.

[0013]

[Action]

[At the time of generating a filter] First, a noise teacher signal N1
(T), N2 (t),... And the target signal teacher signal S1
(T), S2 (t),... Are subjected to wavelet transform by the wavelet transform means, and their wavelet transform coefficient sequence Ws + N (t, a) is obtained.

Similarly, the teacher signal S1 of the target signal
(T), S2 (t),... Are wavelet-transformed by the above-mentioned wavelet transform means to obtain a wavelet transform coefficient sequence WS (t, a) of only the teacher signal of the target signal.

The filter generation means uses, for example, a linear filter with a time delay, and outputs the wavelet transform coefficient sequence Ws + N (t, a) to the wavelet transform of only the teacher signal of the target signal. Coefficient sequence Ws
Find a filter coefficient that is closest to (t, a),
Determine the filter.

[At the time of filtering] When filtering the mixed signal V (t) from a sensor or the like, a wavelet transform is performed by a wavelet transform means as preprocessing to obtain a wavelet transform coefficient sequence WV (t, a).

The wavelet transform coefficient sequence W
Respect V (t, a), by the filtering means performs filtering to remove the wavelet coefficients of noise to obtain a wavelet transform coefficient ^{W f V (t, a)} .

Further, as post-processing, the filtered wavelet transform coefficient W ^f V (t, a) is inversely transformed by wavelet inverse transform means to obtain a noise-free signal W ^f (t).

As described above, a wavelet transform is introduced as a pre-process of filtering, and a wavelet inverse transform is introduced as a post-process, and filtering is performed in the form of a wavelet coefficient. Can be performed in a time-frequency space. The wavelet coefficients include information contained in the original signal and are decomposed in the frequency axis direction while keeping the time axis, so that a higher degree of separation can be realized as compared to filtering on the original signal.

[0020]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the overall configuration of the device according to the present invention. As an example of the present apparatus, a signal processing apparatus that removes a noise signal from a sensor signal and extracts only a signal of an extraction target will be described.

In FIG. 1, reference numeral 1 denotes a signal processing device for performing a filter operation for generating a signal obtained by removing only a noise component from a mixed signal V (t) in a time series in which noise and a target signal are mixed. The signal processing device 1 includes a wavelet coefficient filter generation device 2 and a wavelet filter device 3 that removes noise of a time-series mixed signal V (t) using the filter generated by the wavelet coefficient filter generation device 2. Has become.

The wavelet coefficient filter generator 2 comprises a wavelet transformer 4 and a filter generator 5, and includes a noise teacher signal Ni (t) (i is an arbitrary number of 1 or more) and a target signal teacher signal Sj. (T) (j
Is given by one or more), the wavelet transform coefficient sequences of all the teacher signals obtained by using the wavelet transform device 4 are represented by WS + N (t, a), (a: scale, t: Assuming that the wavelet transform coefficient sequence of only the teacher signal of the target signal is WS (t, a), when the wavelet transform coefficient sequence WS + N (t, a) is input, the wavelet transform coefficient sequence WS (t, a) ) To generate a coefficient sequence closest to the wavelet coefficient filter F
(W) is generated by the filter generation device 5.

The wavelet filter device 3 comprises a wavelet transform device 4 ', a filtering device 6 and an inverse wavelet transform device 7. When a time-series signal V (t) in which noise and a target signal are mixed is supplied, the wavelet filter device 3' Mixed signal V obtained by converter 4 '
In the wavelet transform coefficient sequence WV (t, a) of (t),
The filter F (W) obtained by the wavelet coefficient filter generation device 2 is applied by the filtering device 6, and
A coefficient sequence W ^f V (t, a) = F (W V (t, a)) is generated, and the W ^f V (t, a) is subjected to inverse wavelet transform by the inverse wavelet transform device 7 to generate a signal V ^f
By obtaining (t), noise components are removed from the time-series mixed signal V (t), which is the original signal. In FIG. 1,
Although two wavelet transform devices 4 and 4 'are shown for simplicity of explanation, one device is actually used many times.

Next, the filter generation processing and the filtering processing in the above embodiment will be described with reference to the flowcharts of FIGS. [Generation of Filter] (I) As shown in FIGS. 2 and 3, the noise teacher signal N
1 (t), N2 (t),... And the target signal S
1 (t), S2 (t),...
To obtain a wavelet transform coefficient sequence Ws + N (t, a) (step A1).
).

Here, t is time, and when the sum of all teacher signal lengths is T, the coefficient sequence is defined as 0 <t ≦ T.
a is a, where s is the scale of the wavelet transform.
= Log ₂ s. Wavelet transform coefficient sequence W
s + N (t, a) can be defined as -∞ <a <∞, of which a set of values of a discretely selected to be sufficient for the inverse transformation, {ai | i = 1 to A} (For example, ai = i when an orthogonal basis is used as the wavelet transform basis).

(II) Similarly, the teacher signal S1 of the target signal
(T), S2 (t),... Are wavelet-transformed by the wavelet transform unit 4 to obtain a wavelet transform coefficient sequence WS (t, a) of only the teacher signal of the target signal (step A2).

In this case, the range of a is {ai | i = 1 to A}, which is the same as the above (I). For t,
By assigning 0 to the interval of the noise teacher signal of (I), 0 <t ≦ T is also set similarly to (I).

(III) Filter Generation Apparatus 5 The filter generation apparatus 5 includes a wavelet transform coefficient sequence Ws +
When N (t, a) is input, a filter coefficient is determined such that the output E is closest to the wavelet transform coefficient sequence Ws (t, a) of only the teacher signal of the target signal, and a filter is determined.

For example, using a linear filter with a time delay, the output E to the wavelet transform coefficient sequence Ws + N (t, a) and the wavelet transform coefficient sequence Ws (t, a)
The coefficient is determined so that the square error of is minimized (step A3).

[0030] Specifically, the wavelet transform coefficients WS + N of the input signal (t, a), when the filtered wavelet transform coefficients ^{W f S + N (t,} a), and, the wavelet transform coefficients W ^f S + N (t, a)
｛W (i, j) | t−d ≦ i ≦ t + d, j = a1,..., AA
(D is a time delay width and is a constant). This is shown in the following equation

[0031]

(Equation 1) _Is determined by determining the weights w _ij of.

That is, the wavelet transform coefficient sequence WS + N (t, a) output from the wavelet transform device 4
When was input to the filter, the output E is as closest to WS (t, a), ^{i.e., E = | W f S +} N (t, a) -WS (t, a) | is minimum Such a weight w _ij may be obtained. This is a least squares method, and weights w _ij are obtained algebraically.

[Filtering] (i) When filtering a mixed signal V (t) from a sensor or the like, it is converted as a pre-process by a wavelet transform device 4 'as shown in FIG. 2, and a wavelet transform coefficient sequence WV (t, a) is obtained (step A4). here,
t is 1 to signal length, and the range of a is {ai | i = 1 to A}. (Ii) Then, this wavelet transform coefficient sequence WV
Using the weight w _ij obtained in the above (III) for (t, a),

[0034]

(Equation 2) Performs filtering of the wavelet transform (step A5).

(Iii) Further, as post-processing, the above (i)
Wavelet transform coefficient W ^f filtered in i)
V (t, a) is inversely transformed by the inverse wavelet transform unit 7 to obtain a noise-free signal ^Wf (t) (step A6).

As described above, the wavelet transform is introduced as the pre-processing of the filtering, and the inverse wavelet transform is introduced as the post-processing, and the filtering is performed in the form of the wavelet coefficients. In a time-frequency space. In FIG. 4, a shown on the vertical axis is the logarithm of the scale of the wavelet transform, and is associated with the frequency. As can be seen from the fact that the original signal can be restored by the inverse transform, the wavelet coefficients include information included in the original signal,
In addition, since the signal is decomposed in the frequency axis direction while maintaining the time axis, a higher degree of separation can be realized as compared with filtering of the original signal, and the noise component can be reliably removed.

In the above embodiment, a linear filter with a time delay is used as a filter. However, the present invention is not limited to this, and various types of filters can be applied. For example, in a neural network, a wavelet transform coefficient sequence WS + N (t,
Various methods can be used, such as learning to minimize the error between the output when a) is input and the wavelet transform coefficient sequence WS (t, a).

The signal processing device according to the present invention is applied to an abnormality detection device, a monitoring device, an inspection device using a measurement signal such as a monitor signal or a sensor signal, a noise removal device for a sound / voice signal, a signal enhancement device, and the like. can do.

[0039]

As described above in detail, according to the present invention, a wavelet transform is applied as a pre-process for filtering a mixed signal, and a wavelet inverse transform is applied as a post-process to remove noise components in a space of wavelet coefficients. Is performed, the noise and the target signal overlap to deform the signal shape, or the S / N of the measurement signal
Even when it is difficult to remove noise by the conventional method such as when the ratio is small, the noise component can be reliably removed, and a very high filtering ability can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an overall configuration of a signal processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an overall processing operation in the embodiment.

FIG. 3 is a flowchart showing a processing operation of the filter generation device in the embodiment.

FIG. 4 is an exemplary view for explaining a noise separating operation in the embodiment.

FIG. 5 is a diagram showing a configuration example of a conventional signal processing device.

FIG. 6 is a diagram showing another configuration example of a conventional signal processing device.

FIG. 7 is a diagram for explaining a noise separating operation of a conventional signal processing device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Signal processing apparatus 2 Wavelet coefficient filter generation apparatus 3 Wavelet filter apparatus 4, 4 'wavelet transformation apparatus 5 Filter generation apparatus 6 Filtering apparatus 7 Wavelet inverse transformation apparatus

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Shigeatsu Katayama 1-1-1, Wadasaki-cho, Hyogo-ku, Kobe-shi, Hyogo Prefecture Inside Mitsubishi Heavy Industries, Ltd.Kobe Shipyard (72) Inventor Kiichi Sugimoto, Hyogo-ku, Kobe-shi, Hyogo 1-1 1-1 Tazakicho Mitsubishi Heavy Industries, Ltd. Kobe Shipyard (58) Field surveyed (Int. Cl. ⁷ , DB name) G01D 3/00 G01D 5/00

Claims (1)

(57) [Claims] 1. A teacher signal Ni (t) of noise (i is an arbitrary number equal to or greater than 1) and a teacher signal Sj (t) (j of a target signal).
Is given by 1 or more), the wavelet transform coefficient sequence WS + N (t, a) (a: scale, t: time) of all the teacher signals, and only the teacher signal of the target signal Wavelet transform means for obtaining a wavelet transform coefficient sequence WS (t, a) of the following formula: Wavelet transform coefficient sequence WS + N (t, a) output from the wavelet transform means is a wavelet transform coefficient sequence WS (t, a) Filter generation means for generating a wavelet coefficient filter F which comes closest to the following: a time-series mixed signal V (t) in which noise and a target signal are mixed
Is given, the wavelet transform means generates a wavelet transform coefficient sequence WV of the mixed signal V (t).
Means for calculating (t, a); and a wavelet transform coefficient sequence W V (t, a) determined by the wavelet transform means, which is filtered by the filter F determined by the filter generating means to obtain a coefficient sequence W ^f
Filtering means for generating V (t, a); and a coefficient sequence W ^f V generated by the filtering means.
A signal processing apparatus comprising: wavelet inverse transform means for performing a wavelet inverse transform on (t, a) and outputting a noise-removed signal V ^f (t).