JP7577485B2 - Signal processing method, signal processing device and program - Google Patents

Description

The present invention relates to a signal processing method, a signal processing device, and a program.

In the field of signal processing, a process is carried out to remove noise components contained in a measurement signal measured by a measurement sensor.

The signal processing method described in Patent Document 1 performs processing to remove noise components from multiple measurement signals measured in a state in which information signals and noise signals are mixed, and extract information signals (see Patent Document 1). Specifically, the signal processing method acquires multiple measurement signals at different measurement locations, detects correlation between the multiple measurement signals, detects phase synchrony between the multiple measurement signals, and removes uncorrelated components with low correlation and non-phase synchronized components with low phase synchrony from each of the multiple measurement signals, and outputs multiple estimated signals that are estimated to be information signals.

JP 2017-625 A

However, the signal processing method described in Patent Document 1 can remove uncorrelated components with low correlation between multiple measurement signals and non-phase-synchronous components with low phase synchronism as noise components, but cannot remove noise components with other characteristics.

For example, with this signal processing method, it was difficult to remove noise components when the correlation between the signals desired to be acquired among multiple measurement signals was low, or when the phase synchronism between the signals desired to be acquired was low.
For example, in this signal processing method, when the frequency of the signal to be acquired is close to the frequency of the noise components, it is difficult to remove the noise components without affecting the signal to be acquired, even if processing is performed using a filter such as a low-pass filter or a high-pass filter.
For example, in this signal processing method, when there is one measurement signal, it is not possible to remove the noise component.
For example, in the signal processing method, when a noise component superimposed on a measurement signal has correlation and phase synchronization with a component of a signal that is desired to be acquired, it is difficult to remove the noise component.

The present invention has been made in consideration of these circumstances, and aims to provide a signal processing method, signal processing device, and program that can effectively extract noise components superimposed on a measurement signal.

One aspect is a signal processing method in which a computer extracts, as noise components, common components between a measurement signal measured in a state in which a target signal and a noise signal are mixed, and a reference signal measured in the presence of the noise signal. The computer uses a singular value decomposition technique to extract vectors that constitute the measurement signal and the reference signal, detects singular values that are values that represent the magnitude of influence that the vector has on constituting the signal, and selects, from among a plurality of components that are candidates for the noise components, one or more components that correspond to singular values that exceed a predetermined threshold or one or more components that correspond to singular values that are equal to or greater than the predetermined threshold, and extracts the selected components as the noise components .
One aspect is a signal processing method in which a computer extracts, as noise components, common components between a measurement signal measured in a state where a target signal and a noise signal are mixed, and a reference signal measured in the presence of the noise signal, in which the computer uses a singular value decomposition technique to extract vectors that constitute the measurement signal and the reference signal, detects singular values that are values that represent the magnitude of influence that the vector has on forming the signal, selects one or more components selected by the user based on the content of the operation performed by the user from among a plurality of components that are candidates for the noise component, and extracts the selected components as the noise components.

One aspect is a signal processing device that includes a measurement signal acquisition unit that acquires a measurement signal measured in a state where a target signal and a noise signal are mixed, a reference signal acquisition unit that acquires a reference signal measured in the presence of the noise signal, and a signal processing unit that extracts common components between the measurement signal acquired by the measurement signal acquisition unit and the reference signal acquired by the reference signal acquisition unit as noise components, wherein the signal processing unit uses a singular value decomposition technique to extract vectors that constitute the measurement signal and the reference signal, detects singular values that are values that represent the magnitude of influence that the vector has on constituting a signal, and selects one or more components that correspond to singular values that exceed a predetermined threshold or one or more components that correspond to singular values that are equal to or greater than the predetermined threshold from a plurality of components that are candidates for the noise components, and extracts the selected components as the noise components .
One aspect is a signal processing device comprising: a measurement signal acquisition unit that acquires a measurement signal measured in a state where a target signal and a noise signal are mixed; a reference signal acquisition unit that acquires a reference signal measured in the presence of the noise signal; and a signal processing unit that extracts common components between the measurement signal acquired by the measurement signal acquisition unit and the reference signal acquired by the reference signal acquisition unit as noise components, wherein the signal processing unit uses a singular value decomposition technique to extract vectors that constitute the measurement signal and the reference signal, detects singular values that are values that represent the magnitude of influence that the vector has on constructing a signal, selects one or more components selected by the user based on the content of an operation performed by the user from a plurality of components that are candidates for the noise component, and extracts the selected components as the noise components.

One aspect is a program for causing a computer to execute a first step of acquiring a measurement signal measured in a state in which a target signal and a noise signal are mixed, a second step of acquiring a reference signal measured in a state in which the noise signal is present, and a third step of extracting common components between the acquired measurement signal and the acquired reference signal as noise components , in which the third step uses a singular value decomposition technique to extract vectors constituting the measurement signal and the reference signal, detect singular values which are values representing the magnitude of influence that the vector has on constituting the signal, select one or more components corresponding to singular values exceeding a predetermined threshold or one or more components corresponding to singular values equal to or greater than the predetermined threshold from among a plurality of components that are candidates for the noise components, and extract the selected components as the noise components .
One aspect is a program for causing a computer to execute a first step of acquiring a measurement signal measured in a state in which a target signal and a noise signal are mixed, a second step of acquiring a reference signal measured in a state in which the noise signal is present, and a third step of extracting common components between the acquired measurement signal and the acquired reference signal as noise components, wherein the third step uses a singular value decomposition technique to extract vectors constituting the measurement signal and the reference signal, detects singular values which are values representing the magnitude of influence that the vector has on constituting the signal, selects one or more components selected by the user based on the content of an operation performed by the user from among a plurality of components that are candidates for the noise components, and extracts the selected components as the noise components.

According to the present invention, a signal processing method, a signal processing device, and a program can effectively extract noise components superimposed on a measurement signal.

1 is a diagram showing a schematic configuration of a signal processing system according to an embodiment (first embodiment). FIG. 2 is a diagram showing an example of a procedure of processing performed in a signal processing device according to an embodiment. 5 is a diagram showing an example of a procedure of a target signal extraction process performed in the signal processing device according to the embodiment; FIG. FIG. 1 is a diagram for explaining the principle of ANC. FIG. 1 is a diagram for explaining the principle of ANC. FIG. 4 is a diagram illustrating an example of a measurement signal. FIG. 4 is a diagram illustrating an example of a reference signal. FIG. 4 is a diagram illustrating an example of a target signal. 1 is a diagram showing an example of a procedure of a noise component extraction process using singular value decomposition performed in a signal processing device according to an embodiment. FIG. 1 illustrates the principle of singular value decomposition. 1A to 1E are diagrams illustrating examples of reference signals. 13A to 13F are diagrams showing examples of components that are candidates for noise components. FIG. 13 is a diagram illustrating an example of a singular value. 1 is a diagram showing an example of a procedure of a noise component extraction process using principal component analysis performed in a signal processing device according to an embodiment. 11 is a diagram showing an example of a procedure of a noise component extraction process using factor analysis performed in the signal processing device according to the embodiment. FIG. FIG. 11 is a diagram showing a schematic configuration of a signal processing system according to an embodiment (second embodiment).

The following describes an embodiment of the present invention with reference to the drawings.

First Embodiment
<Signal Processing System>
FIG. 1 is a diagram showing a schematic configuration of a signal processing system 1 according to an embodiment (first embodiment).
The signal processing system 1 according to this embodiment includes a signal processing device 11, one measurement sensor 21, and M reference sensors 22-1 to 22-M.
The signal processing device 11 includes a measurement signal acquisition unit 31, a reference signal acquisition unit 32, a signal processing unit 33, and an output unit 34.
The signal processing unit 33 includes a noise component extraction unit 41 and a target signal extraction unit 42 .

The signal processing device 11 may be, for example, a computer including a processor such as a CPU (Central Processing Unit) and a storage unit such as a RAM (Random Access Memory) and a ROM (Read Only Memory). In this case, the signal processing device 11 may execute various processes by executing a program stored in the storage unit with the processor. Furthermore, the signal processing device 11 may use various parameters stored in the storage unit when executing a program with the processor.

In this embodiment, we will explain the case where M, the number of reference sensors 22-1 to 22-M, is an integer greater than or equal to 2, but M may also be 1, that is, the signal processing system 1 may be configured to include one reference sensor 22-1.
In this embodiment, for convenience of explanation, a case where the signal processing system 1 includes one reference sensor 22-1 will also be described.
Although the signal processing system 1 includes a plurality of reference sensors 22-1 to 22-M, when only one reference sensor 22-1 is used, processing is performed similarly to when the signal processing system 1 includes one reference sensor 22-1.

In this embodiment, the measurement sensor 21 is separate from the signal processing device 11, but as another configuration example, the measurement sensor 21 may be provided integrally with the signal processing device 11.
In this embodiment, a configuration is shown in which the reference sensors 22-1 to 22-M are separate from the signal processing device 11. However, as another configuration example, some or all of the reference sensors 22-1 to 22-M may be provided integrally with the signal processing device 11.

The measurement sensor 21 is installed to measure a measurement signal including a target signal. In this embodiment, a signal that is desired to be acquired is referred to as a target signal. In this embodiment, a signal measured by the measurement sensor 21 is referred to as a measurement signal.
The measurement signal includes a target signal and a noise component. In this embodiment, signals other than the target signal are referred to as noise signals. In addition, in this embodiment, signal components of the noise signal are referred to as noise components.

Each of the reference sensors 22-1 to 22-M is provided to measure a reference signal including a noise signal. In this embodiment, the signals measured by each of the reference sensors 22-1 to 22-M are referred to as reference signals.
In this embodiment, the noise signals contained in the reference signals measured by the respective reference sensors 22-1 to 22-M are caused by the same noise source as the noise components contained in the measurement signals.

The noise source may be an artificial environment caused by the device, or may be a natural environment other than the device, such as the earth.
The number of noise sources may be one or more. Here, if the accuracy of removing noise components is affected by the number of reference sensors 22-1 to 22-M relative to the number of noise sources, it is preferable to provide a number of reference sensors 22-1 to 22-M so as to ensure the required accuracy. As an example, it is preferable to provide the same number of reference sensors 22-1 to 22-M as the number of noise sources or a number greater than the number of noise sources.
In addition, when the same noise source generates two or more types of noise signals with different properties, each type of noise signal may be considered to exist as a different noise source.
The property may be frequency. In this case, when the same noise source generates noise signals of two or more different frequencies, it may be considered that a different noise source exists for each noise signal of each frequency.

The measurement sensor 21 is disposed in such a manner that it can measure the target signal. Ideally, the measurement result does not include any signal other than the target signal (signal referred to as a noise signal in this embodiment), but in reality, signals other than the target signal are often included. In this embodiment, it is assumed that the measurement signal includes signal components other than the target signal (components referred to as noise components in this embodiment), and a process for removing the noise components will be described.
Each of the reference sensors 22-1 to 22-M is disposed in such a manner that the noise signal can be significantly measured. Here, in this embodiment, measuring the noise signal significantly means that, among the noise signal and other signals (called target signals in this embodiment) included in the measurement result, the components of the other signals are absent or at the level of error. Although it is a preferable example that the reference signal includes only the noise signal, it may also include components of the noise signal and the target signal. For example, the reference signal may include a component of the target signal to the extent that it is considered to be an error.

The reference signals measured by the respective reference sensors 22-1 to 22-M may or may not include a noise signal caused by the same noise source. The noise signal is a noise signal superimposed on the measurement signal measured by the measurement sensor 21.
The reference signals measured by the respective reference sensors 22-1 to 22-M may or may not include noise signals caused by different noise sources. Note that the noise signals are noise signals superimposed on the measurement signal measured by the measurement sensor 21.
Due to the influence of a noise signal generated from the same noise source, the noise component contained in the measurement signal and the noise signal contained in the reference signal may have different amplitudes or the same amplitudes.

The measurement sensor 21 and each of the reference sensors 22-1 to 22-M are installed, for example, in different manners, so that the measurement sensor 21 measures a target signal and each of the reference sensors 22-1 to 22-M significantly measures a noise signal.
The manner in which the measurement sensor 21 is installed and the manner in which each of the reference sensors 22-1 to 22-M is installed may be adjusted, for example, in terms of the location of installation or the attitude of installation.
For example, the measurement sensor 21 may be installed near the source of the target signal, and the reference sensors 22-1 to 22-M may be installed at locations far enough away from the source that the target signal is not significantly measured. In this embodiment, the fact that the target signal is not significantly measured means that a noise signal is significantly measured.

The measurement sensor 21 and each of the reference sensors 22-1 to 22-M may be sensors having the same performance or characteristics, or may be sensors having different performance or characteristics. The performance or the like may be the sensitivity of the measurement (the degree of intensity of the signal obtained by the measurement).
The reference sensors 22-1 to 22-M may be sensors having the same performance or characteristics, or may be sensors having different performance or characteristics. The performance or the like may be the measurement sensitivity (the degree of intensity of the signal obtained by the measurement) or the frequency that can be measured as the measurement target (the frequency of the signal to be measured).

The processing performed in the signal processing system 1 will now be described.
The measurement sensor 21 measures a measurement signal along the progression of time. The measurement signal along the progression of time may be a time-continuous signal, or may be a signal obtained at predetermined discrete sampling timings such as at a constant period.
The measurement sensor 21 outputs a measured signal to the signal processing device 11 .

Each of the reference sensors 22-1 to 22-M measures a reference signal along the progression of time. The reference signal along the progression of time may be a time-continuous signal, or may be a signal obtained at a predetermined discrete sampling timing such as a fixed period.
Each of the reference sensors 22 - 1 to 22 -M outputs a measured reference signal to the signal processing device 11 .

In this embodiment, the process of measuring the measurement signal by the measurement sensor 21 and the process of acquiring the reference signals by each of the reference sensors 22-1 to 22-M are performed simultaneously in parallel in the same time period. The measurement signal and the reference signal are aligned on the time axis in the same time period. In this embodiment, the measurement signal and the reference signal are aligned on the time axis in the same time period means that the measurement signal and the reference signal are signals measured from the start time to the end time of the same time period, and signal portions measured at the same time can be processed in correspondence with each other (in this embodiment, noise component extraction processing and removal processing, etc.).
As another configuration example, the measurement start timing or measurement end timing of the measurement sensor 21 and the multiple reference sensors 22-1 to 22-M do not necessarily have to be the same, and there may be a common time period during which measurements are performed by all the sensors. In this case, the common time period is used as the same time period described above to perform processing (in this embodiment, the noise component extraction processing and removal processing, etc.), and the signal portion at times other than the common time period may not be used for processing. In other words, in this case, the common time period may be used as the same time period when the measurement signal and the multiple reference signals are measured during the same time period described above.

In this embodiment, the measurement sensor 21 and each of the reference sensors 22-1 to 22-M are magnetic sensors that measure magnetic signals.
The measurement sensor 21 and each of the reference sensors 22-1 to 22-M may be any sensor, such as an electric field sensor that measures an electric field signal, a current sensor that detects a current, a voltage sensor that measures a voltage, a sound sensor that measures a sound, a temperature sensor that detects a temperature, a pressure sensor that measures a pressure, an optical sensor that measures light, an electromagnetic wave sensor that measures an electromagnetic wave, etc. The electromagnetic wave sensor may be, for example, an infrared sensor that measures infrared rays, or a radio wave sensor that measures radio waves.

The measurement signal acquiring unit 31 inputs and acquires the measurement signal output from the measurement sensor 21 , and outputs the measurement signal to the signal processing unit 33 .
The reference signal acquisition unit 32 inputs and acquires the reference signals output from the respective reference sensors 22 - 1 to 22 -M, and outputs the acquired reference signals to the signal processing unit 33 .

The noise component extraction unit 41 extracts common components of the measurement signal acquired by the measurement signal acquisition unit 31 and the reference signal acquired by the reference signal acquisition unit 32 as noise components contained in the measurement signal. The noise components are signal components other than the target signal contained in the measurement signal.
The common component may be, for example, a signal component having a waveform similar to that of the measurement signal and the reference signal.

The target signal extraction unit 42 removes the noise components extracted by the noise component extraction unit 41 from the measurement signal acquired by the measurement signal acquisition unit 31, and extracts the signal from which the noise components have been removed as the target signal from the measurement signal.
The target signal extraction unit 42 removes the noise component from the measurement signal, for example, by aligning the time axis of the measurement signal and the noise component and subtracting the noise component from the measurement signal.

In this embodiment, the target signal extraction unit 42 corrects the noise components extracted by the noise component extraction unit 41 based on the correlation between the measurement signal and the noise components, and removes the corrected noise components from the measurement signal. The correction adjusts the amplitude of the noise components before correction according to the measurement signal, thereby making it possible to more effectively remove noise. The process of adjusting the amplitude of the noise components may be a process of multiplying the amplitude by a predetermined number while maintaining the waveform of the noise components. Here, instead of correcting the amplitude, a correction of the intensity may be performed.
If such a correction is not required, the correction need not be performed.

Here, the target signal extracted by the target signal extraction unit 42 ideally does not contain any signal components other than the target signal, but in reality, depending on the accuracy of the noise removal, it may contain signal components other than the target signal (noise components that could not be completely removed in this embodiment).

The signal processing unit 33 outputs the target signal extracted by the target signal extraction unit 42 to the output unit 34 .
The output unit 34 outputs information related to the target signal output from the signal processing unit 33. For example, the output unit 34 may output the target signal to an external device, or may display information related to the target signal on a screen.

Here, the noise component extraction unit 41 and the target signal extraction unit 42 perform processing while ensuring phase synchronism with each other, with the time axes of the measurement signal and the reference signal aligned.

FIG. 2 is a diagram showing an example of a procedure of processing performed in the signal processing device 11 according to the embodiment.
The processing procedure performed in the signal processing device 11 will be described as steps S1 to S4.

(Step S1)
The measurement signal acquisition unit 31 and the reference signal acquisition unit 32 acquire a measurement signal and a reference signal.

(Step S2)
The noise component extractor 41 extracts the noise component based on the measurement signal and the reference signal.

(Step S3)
The target signal extractor 42 removes noise components from the measurement signal and extracts the target signal.

(Step S4)
The output unit 34 outputs the extracted target signal.
The output unit 34 may output other information.

FIG. 3 is a diagram showing an example of a procedure of a target signal extraction process performed in the signal processing device 11 according to the embodiment.
An example of the procedure of the process of step S3 shown in FIG. 2 will be described as steps S21 to S22.

(Step S21)
The target signal extractor 42 determines the corrected noise components based on the correlation between the measurement signal and the noise components extracted by the noise component extractor 41 .

(Step S22)
The target signal extraction unit 42 extracts the target signal by removing the corrected noise components from the measurement signal.
In addition, if correction of the noise components extracted by the noise component extraction unit 41 is not necessary, the correction may not be performed, and, for example, the noise components before and after correction may be considered to be the same.

<ANC>
The signal processing device 11 may use ANC (Adaptive Noise Cancelling) to remove noise components from the measurement signal.
4 and 5 are diagrams for explaining the principle of ANC.
4 and 5, the horizontal axis represents time, and the vertical axis represents amplitude. The amplitude may be obtained in the form of voltage or the like. Note that the scales on the horizontal and vertical axes are omitted.
The graphs in FIG. 4 and FIG. 5 show the case where the time periods on the horizontal axis are aligned.
4 shows an outline of the waveform of the measurement signal 1011. The measurement signal 1011 includes a target signal 1021 and a noise component 1022 due to a noise signal.
FIG. 5 shows an outline of the waveform of the reference signal 1031.

For the sake of simplicity, it is assumed that reference signal 1031 is only a noise signal, and the noise component contained in measurement signal 1011 is a signal component that is correlated with reference signal 1031 .
In this case, the target signal 1021 corresponds to the result of subtracting a signal that is a predetermined number times the reference signal 1031 from the measurement signal 1011. The value of the predetermined number times may be obtained by any analysis method.
In this case, the waveform of the noise component contained in the measurement signal 1011 and the waveform of the reference signal 1031 are similar to each other.
Here, it is assumed that phase synchronism between measurement signal 1011 and reference signal 1031 is ensured.

FIG. 6 is a diagram showing an example of the measurement signal 1211.
7 is a diagram showing an example of the reference signal 1221. In this example, the reference signal 1221 is only a noise signal.
FIG. 8 is a diagram showing an example of the target signal 1231. As shown in FIG.
6 to 8, the horizontal axis represents time, and the vertical axis represents amplitude. The amplitude may be obtained in the form of voltage or the like. Note that the scales on the horizontal and vertical axes are omitted.
The graphs in FIG. 6 to FIG. 8 show cases where the time periods on the horizontal axis are aligned.

<Signal processing timing>
In this embodiment, the signal processing device 11 temporarily stores the measurement signal acquired by the measurement signal acquisition unit 31 and the reference signal acquired by the reference signal acquisition unit 32 in a memory unit, and then uses the stored measurement signal and reference signal to perform noise component extraction processing and noise component removal processing by the signal processing unit 33.
As described above, in this embodiment, the processing performed by the signal processing unit 33 is not real-time processing, but as another configuration example, the signal processing device 11 may perform real-time signal processing by the signal processing unit 33.

<Specific example of noise component extraction processing>
As specific examples of the noise component extraction process performed by the noise component extraction unit 41, an example using singular value decomposition, an example using principal component analysis, and an example using factor analysis will be shown.

[Example using singular value decomposition]
FIG. 9 is a diagram showing an example of a procedure of a noise component extraction process using singular value decomposition performed in the signal processing device 11 according to the embodiment.
An example of the procedure of the process of step S2 shown in FIG. 2 will be described as steps S31 to S32.

(Step S31)
The noise component extractor 41 uses a singular value decomposition technique to extract vectors that make up the signal based on the measurement signal and the reference signal, and detects a value that represents the magnitude of the influence that the vector has on the composition of the signal. The value corresponds to a singular value.

(Step S32)
The noise component extracting section 41 extracts vectors that have a large effect on the composition of the signal, that is, vectors with high singular values, as noise components.

The noise component extraction process using singular value decomposition will be described more specifically with reference to FIGS.
FIG. 10 is a diagram illustrating the principle of singular value decomposition.

[Equation 1]
A=UΣV ^T ... (Formula 1)

Equation 1 is an example of an equation used in singular value decomposition.
A, U, Σ, and V each represent a matrix.
FIG. 10 shows an outline of A, U, Σ, and V.
For example, U is a left singular vector, Σ is a singular value matrix, and V is a right singular vector.

In matrix A, signals 1 to p correspond to a plurality of p rows, respectively, and dimensions 1 to q correspond to a plurality of q columns, respectively.
In this embodiment, a measurement signal is used as signal 1, and (p-1) reference signals are used as signals 2 to p. The number of reference signals, (p-1), may be 2 or more, or may be 1.
The order of signals 1 to p may be arbitrary.
Dimensions 1 to q represent time components. That is, dimensions 1 to q of signal 1 are q different time components of signal 1, and similarly, dimensions 1 to q of signal 2 to signal p. In this embodiment, dimensions 1 to q are arranged in line with the progression of time.

In singular value decomposition, once matrix A is determined, matrix U, matrix Σ, and matrix V are also determined.
For example, matrix A is a (p×q) matrix, matrix U is a (p×n) matrix, matrix Σ is an (n×n) matrix, and matrix V is an (n×q) matrix, where n is an integer.
Note that the numbers of rows and columns of these matrices are merely examples and are not limited to these.

In the matrix Σ, the singular values σ1 to σn of the matrix A are arranged as diagonal elements. The singular values σ1 to σn are arranged in order from largest to smallest.
The larger the singular value, the higher the degree of correlation in all of signals 1 to p, i.e., the component has a large influence on the composition of the signal. In this embodiment, such a component that has a large influence on the composition of the signal is considered to have a high possibility of being a noise component.
By selecting one or more of the larger singular values from among the singular values σ1 to σn, a vector representation with reduced dimensions is obtained for the matrix U. Using this, in this embodiment, one or more components corresponding to the larger singular values are selected from a plurality of components that may be noise components (candidate noise components), and the selected components are regarded as noise components and removed from the measurement signal.
Here, when p signals, signal 1 to signal p, are used, p or less candidates for noise components are obtained.

A case in which a measurement signal and five reference signals are used will be described with reference to FIGS.
11A to 11E are diagrams showing examples of five reference signals 2011 to 2015.
11A to 11E, the horizontal axis represents time, and the vertical axis represents amplitude. The amplitude may be obtained in the form of voltage or the like. Note that the scales on the horizontal and vertical axes are omitted.
The graphs in FIGS. 11A to 11E show cases where the time periods on the horizontal axis are aligned.
11A to 11E show reference signals 2011 to 2015 measured by the five reference sensors 22-1 to 22-5, respectively.

12A to 12F are diagrams showing examples of components that are candidates for noise components. These components will be referred to as candidate noise components in the following description.
12A to 12F, the horizontal axis represents time, and the vertical axis represents amplitude. The amplitude may be obtained in the form of voltage or the like. Note that the scales on the horizontal and vertical axes are omitted.
The graphs in FIGS. 12(A) to (F) show cases where the time periods on the horizontal axis are aligned.
12A to 12F show six noise candidate components 2021 to 2026 extracted by singular value decomposition using the measurement signal and reference signals 2011 to 2015.

FIG. 13 is a diagram showing an example of the characteristic 2031 of the singular value.
13, the horizontal axis represents the component number, and the vertical axis represents the singular value. The component numbers are the numbers of the six noise candidate components 2021 to 2026, and are assigned numbers 1 to 6 in order.

The values decrease in the order from the largest singular value σ1 to the largest singular value σ6.
For example, when extracting noise components corresponding to two singular values, a noise candidate component 2021 corresponding to the singular value σ1 and a noise candidate component 2022 corresponding to the singular value σ2 are selected and extracted as noise components.
In addition, when extracting noise components corresponding to one singular value, or when extracting noise components corresponding to three or more singular values, the number of noise candidate components corresponding to the top singular values is selected and extracted as noise components.

Here, any method may be used to select one or more noise candidate components as noise components from among a plurality of noise candidate components.
For example, in the signal processing device 11, a threshold value for singular values may be stored in a memory unit, and the noise component extraction unit 41 may select a noise candidate component corresponding to a singular value that exceeds the threshold value (or is equal to or greater than the threshold value) as a noise component.
For example, a method may be used in which a predetermined number of singular values is stored in a storage unit in the signal processing device 11, and the noise component extraction unit 41 selects the predetermined number of singular values in ascending order of the singular values, and selects the noise candidate components corresponding to the selected singular values as noise components. In this case, when the number of noise components predicted to be included in the measurement signal is known or estimated in advance, the number may be set as the predetermined number.

For example, the signal processing device 11 may receive the contents of an operation performed by a user (person), and select one or more noise candidate components as noise components from among a plurality of noise candidate components based on the contents of the operation. In other words, the user may select the components to be noise components.

A known algorithm may be used as the singular value decomposition algorithm. In this case, the signal processing device 11 stores the singular value decomposition algorithm and calculates and obtains singular values, etc., by substituting parameter values related to the measurement signal and the reference signal into the algorithm.

[Example of using principal component analysis]
FIG. 14 is a diagram showing an example of a procedure of a noise component extraction process using principal component analysis performed in the signal processing device 11 according to the embodiment.
An example of the procedure of the process of step S2 shown in FIG. 2 will be described as steps S41 to S42.

(Step S41)
The noise component extractor 41 performs dimension reduction based on the measurement signal and the reference signal using a principal component analysis technique.

(Step S42)
The noise component extractor 41 extracts the principal components as noise components.

In this embodiment, in the principal component analysis, components that are included in one measurement signal and one or more reference signals are identified as principal components.
In general, principal component analysis involves synthesizing individual ratings and summarizing them into principal components, and then explaining each principal component using variables.

A known algorithm may be used as the algorithm for the principal component analysis. In this case, the signal processing device 11 stores the algorithm for the principal component analysis and calculates and acquires the principal components by substituting the parameter values related to the measurement signal and the reference signal into the algorithm.
Furthermore, in the signal processing device 11, a condition for determining a component to be removed from the measurement signal as a noise component based on the result of the principal component analysis may be set in advance. The condition may be stored in a storage unit of the signal processing device 11.

[Example of using factor analysis]
FIG. 15 is a diagram showing an example of a procedure of a noise component extraction process using factor analysis performed in the signal processing device 11 according to the embodiment.
An example of the procedure of the process of step S2 shown in FIG. 2 will be described as steps S51 to S52.

(Step S51)
The noise component extractor 41 detects fluctuations based on the measurement signal and the reference signal using a factor analysis technique.

(Step S52)
The noise component extracting section 41 extracts the components that have a common fluctuation as noise components.

In this embodiment, in the factor analysis, variations common to all of a measurement signal and one or more reference signals are identified as noise components.
Generally, in factor analysis, individual evaluations are broken down and summarized into factors, and the factors are then broken down into individual variables.

A known algorithm may be used as the factor analysis algorithm. In this case, the signal processing device 11 stores the factor analysis algorithm and calculates and acquires factors by substituting parameter values related to the measurement signal and the reference signal into the algorithm.
Furthermore, in the signal processing device 11, a condition for determining a component to be removed from the measurement signal as a noise component based on the result of the factor analysis may be set in advance. The condition may be stored in a storage unit of the signal processing device 11.

<Regarding the First Embodiment>
As described above, in the signal processing system 1 according to this embodiment, the signal processing device 11 executes the following signal processing method.
The signal processing device 11 extracts the target signal using one measurement signal measured in a state where the target signal and a noise signal are mixed, and one or more reference signals that do not primarily acquire the target signal. In this case, the signal processing device 11 extracts a common component from the measurement signal and the reference signal as a noise component.
Therefore, the signal processing device 11 can effectively extract the noise components superimposed on the measurement signal, and can effectively remove the noise components superimposed on the measurement signal by using such noise components.

The signal processing device 11 extracts the common components between the measurement signal and the reference signal as noise components, and thereby can extract noise components that are superimposed on the measurement signal and have correlation and phase synchronization.
For example, when the measurement signal and the reference signal are measured in different manners, by extracting the common components between the measurement signal and the reference signal as noise components, it is possible to extract noise components that are suitable for the measurement signal, compared to when the reference signal is treated as a noise signal as is.

The signal processing device 11 is capable of extracting noise components superimposed on the measurement signal without using a filter such as a low-pass filter or a high-pass filter.
For example, even when the frequency of the target signal and the frequency of the noise component are close to each other, the signal processing device 11 can extract the noise component without using a filter while suppressing the influence of distortion and the like on the target signal. This allows the signal processing device 11 to extract the noise component from the measurement signal while suppressing the influence of distortion and the like on the extracted target signal.
The signal processing device 11 may use a filter in the process of extracting noise components superimposed on the measurement signal.

Even when there is only one measurement signal, the signal processing device 11 can extract a noise component from the measurement signal. As a result, even when there is only one measurement signal, the signal processing device 11 can remove a noise component from the measurement signal.
The signal processing device 11 can extract the noise components superimposed on the measurement signal even if the noise components have correlation and phase synchronization with the components of the target signal.

In the signal processing system 1 according to this embodiment, the measurement signal and the reference signal are signals with the same time axis.
Therefore, the signal processing device 11 can easily extract the common portion between the measurement signal and the reference signal as a noise component with high accuracy.

In the signal processing system 1 according to the present embodiment, the measurement signal and the reference signal are signals measured in the same time period, which means that there is a high possibility that the measurement signal and the reference signal contain noise components generated at the same time by the same source.
Therefore, the signal processing device 11 can accurately extract the common portion between the measurement signal and the reference signal as a noise component.

In the signal processing system 1 according to this embodiment, the signal processing device 11 extracts the target signal by calculating the correlation between the extracted noise components and the measurement signal, correcting the noise components based on the result, and removing the corrected noise components from the measurement signal.
Therefore, even if there is variation in amplitude between the extracted noise component and the noise component contained in the measurement signal, the signal processing device 11 can correct the extracted noise component to be closer to the noise component contained in the measurement signal by taking the correlation between them.

In the signal processing system 1 according to this embodiment, the signal processing device 11 uses singular value decomposition as a method for extracting noise components.
Therefore, the signal processing device 11 uses singular value decomposition to extract vectors that constitute a signal, and can extract vectors that have a large influence on the composition of the signal as noise components.

In the signal processing system 1 according to this embodiment, the signal processing device 11 uses principal component analysis as a method for extracting noise components.
Therefore, the signal processing device 11 can use principal component analysis to summarize many quantitative explanatory variables into fewer indicators or synthetic variables (perform dimensional reduction) and extract principal components as noise components.

In the signal processing system 1 according to this embodiment, the signal processing device 11 uses factor analysis as a method for extracting noise components.
Therefore, the signal processing device 11 can estimate and extract the noise components by using factor analysis to separate all fluctuations in the measurement signal into fluctuations common to all data (noise components) and fluctuations specific to the measurement signal.

In the signal processing system 1 according to this embodiment, for example, the measurement signal and the reference signal are magnetic signals, and the noise components contained in the measurement signal are noise components that are superimposed on the target signal by the environmental magnetic field.
For example, even when the target signal is a weak magnetic signal, the signal processing device 11 can remove the influence of the environmental magnetic field from the measurement signal and extract the target signal.

In addition, instead of the word measurement, it may be called measurement or detection, and such cases are also included in this embodiment.

Second Embodiment
FIG. 16 is a diagram showing a schematic configuration of a signal processing system 101 according to an embodiment (second embodiment).
The configuration and operation of the signal processing system 101 are similar to those of the signal processing system 1 shown in FIG. 1 relating to the first embodiment, except that the signal processing system 101 is provided with a plurality of N measurement sensors 21-1 to 21-N, and that the signal processing device 11 processes a plurality of measurement signals.

In this embodiment, for convenience of explanation, the signal processing device 11, each unit of the signal processing device 11, and the reference sensors 22-1 to 22-M will be described using the same reference numerals as those shown in FIG.
In this embodiment, differences from the first embodiment will be mainly described, and a description of similarities with the first embodiment will be omitted.

Each of the measurement sensors 21 - 1 to 21 -N measures a measurement signal and outputs the measurement signal to the signal processing device 11 .
The signal processing device 11 acquires, by the measurement signal acquiring section 31, the measurement signals output from the respective measurement sensors 21-1 to 21-N.
The signal processing device 11 performs a noise component extraction process and a target signal extraction process (noise component removal process) for each measurement signal using the noise component extraction section 41 and the target signal extraction section 42.
The signal processing device 11 outputs, via the output section 34, one or more target signals among a plurality of target signals extracted from a plurality of measurement signals.

As a configuration example different from this embodiment, instead of the multiple measurement sensors 21-1 to 21-N, a single measurement sensor may be provided, and the installation mode of the single measurement sensor may be changed to multiple (e.g., N types). Specifically, by repeatedly performing a process of acquiring a measurement signal and removing noise components from the measurement signal with the installation mode of the single measurement sensor set to a desired mode while changing the mode, it is possible to achieve a process of acquiring a measurement signal and removing noise components from the measurement signal for multiple different installation modes as a whole.

When processing multiple measurement signals, the signal processing unit 33 may process the multiple measurement signals in parallel, or may process the multiple measurement signals in a time-division manner, with each measurement signal being processed in turn.

<Regarding the Second Embodiment>
As described above, in the signal processing system 1 according to this embodiment, the signal processing device 11 executes the following signal processing method.
When there are multiple measurement signals, the signal processing device 11 performs noise removal processing for each measurement signal based on a combination of one measurement signal and one or more reference signals, similar to that of the first embodiment. For example, the signal processing device 11 performs processing for each measurement signal in parallel for the multiple measurement signals, or performs the processing repeatedly in a time-division manner.
Therefore, even when there are a plurality of measurement signals, the signal processing device 11 can effectively remove noise components superimposed on the measurement signals.

The signal processing device 11 can remove noise components from each of the measurement signals even when there is low correlation or no phase synchronism between signals that are desired to be acquired among a plurality of measurement signals.
The signal processing device 11 can remove noise components from each of the measurement signals even when there is correlation and phase synchronism between the signals that are desired to be acquired among a plurality of measurement signals.

The signal processing device 11 can also remove noise components that are not uncorrelated components having low correlation between a plurality of measurement signals or non-phase synchronous components having low phase synchronous characteristics.
For example, the signal processing device 11 can also remove noise components that have correlation and phase synchronization between a plurality of measurement signals.

The signal processing device 11 extracts the noise components for each measurement signal, and therefore can effectively remove the noise components from each measurement signal.
In this embodiment, even when there are many measurement signals, it is possible to effectively remove noise components from each measurement signal without increasing the number of reference signals.

<Configuration example related to the above embodiment>
As one configuration example, in the signal processing method (in the above embodiment, the signal processing method executed by the signal processing device 11), a common component between a measurement signal measured in a state where a target signal and a noise signal are mixed and a reference signal measured in the presence of the noise signal is extracted as a noise component.
As one configuration example, the signal processing method corrects the extracted noise component based on the correlation between the noise component and the measurement signal, and removes the corrected noise component from the measurement signal.

As one configuration example, the signal processing method performs signal processing for a common time period in which the measurement signal and the noise signal are aligned on the same time axis, thereby extracting a common component as a noise component.
As one configuration example, in the signal processing method, the measurement signal and the noise signal are signals measured during the same time period from the same measurement start timing to the same measurement end timing, and this same time period corresponds to a common time period.

In one implementation, the signal processing method uses singular value decomposition to extract common components as noise components.
In one implementation, the signal processing method uses principal component analysis to extract common components as noise components.
In one implementation, the signal processing method uses factor analysis to extract common components as noise components.

As an example configuration, in the signal processing method, when there are multiple measurement signals, processing is performed for each measurement signal.
In one embodiment, in the signal processing method, the measurement signal and the reference signal are magnetic signals.

A signal processing device that executes the above-described signal processing method may be implemented.
As one configuration example, the signal processing device 11 includes a measurement signal acquisition unit 31 that acquires a measurement signal measured in a state where a target signal and a noise signal are mixed, a reference signal acquisition unit 32 that acquires a reference signal measured in a state where a noise signal is present, and a signal processing unit 33 that extracts common components between the measurement signal acquired by the measurement signal acquisition unit 31 and the reference signal acquired by the reference signal acquisition unit 32 as noise components.

Here, in the example of Figure 1, a case is shown in which the measurement sensor 21 and the reference sensors 22-1 to 22-M are configured separately from the signal processing device 11, but as another configuration example, a configuration in which one or more of the measurement sensor 21 and the reference sensors 22-1 to 22-M are provided in the signal processing device 11 may be used.
In addition, the example of Figure 16 shows a case where the measurement sensors 21-1 to 21-N and the reference sensors 22-1 to 22-M are configured separately from the signal processing device 11, but as another configuration example, a configuration in which one or more of the measurement sensors 21-1 to 21-N and the reference sensors 22-1 to 22-M are provided in the signal processing device 11 may be used.

A program for executing the above-described signal processing method may be implemented.
As one configuration example, it is a program for causing a computer (for example, a computer constituting the signal processing device 11) to execute the steps of acquiring a measurement signal measured in a state in which a target signal and a noise signal are mixed, acquiring a reference signal measured in the presence of a noise signal, and extracting a common component between the acquired measurement signal and the acquired reference signal as a noise component.

A program for implementing the functions of any component in any device such as the signal processing device 11 described above may be recorded on a computer-readable recording medium, and the program may be read into a computer system and executed. The term "computer system" as used herein includes hardware such as an operating system or peripheral devices. The term "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD (Compact Disc)-ROM (Read Only Memory), as well as storage devices such as hard disks built into computer systems. The term "computer-readable recording medium" also refers to devices that hold a program for a certain period of time, such as volatile memory inside a computer system that is a server or client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. The volatile memory may be, for example, a RAM (Random Access Memory). The recording medium may be, for example, a non-transitory recording medium.

The above-mentioned program may be transmitted from a computer system in which the program is stored in a storage device or the like to another computer system via a transmission medium or by transmission waves in the transmission medium. Here, the "transmission medium" that transmits the program refers to a medium that has a function of transmitting information, such as a network such as the Internet or a communication line such as a telephone line.
The above program may be for implementing some of the above functions. Furthermore, the above program may be a so-called differential file that can implement the above functions in combination with a program already recorded in the computer system. The differential file may be called a differential program.

In addition, the functions of any component in any device such as the signal processing device 11 described above may be realized by a processor. For example, each process in the embodiment may be realized by a processor that operates based on information such as a program and a computer-readable recording medium that stores information such as a program. Here, the functions of each part of the processor may be realized by individual hardware, or the functions of each part may be realized by integrated hardware. For example, the processor may include hardware, and the hardware may include at least one of a circuit that processes digital signals and a circuit that processes analog signals. For example, the processor may be configured using one or more circuit devices mounted on a circuit board, or one or both of one or more circuit elements. An IC (Integrated Circuit) or the like may be used as the circuit device, and a resistor or a capacitor may be used as the circuit element.

Here, the processor may be, for example, a CPU. However, the processor is not limited to a CPU, and various processors such as a GPU (Graphics Processing Unit) or a DSP (Digital Signal Processor) may be used. The processor may be, for example, a hardware circuit using an ASIC (Application Specific Integrated Circuit). The processor may be, for example, a plurality of CPUs, or a hardware circuit using a plurality of ASICs. The processor may be, for example, a combination of a plurality of CPUs and a hardware circuit using a plurality of ASICs. The processor may include, for example, one or more of an amplifier circuit or a filter circuit that processes an analog signal.

The above describes an embodiment of the present invention in detail with reference to the drawings, but the specific configuration is not limited to this embodiment, and includes designs that do not deviate from the gist of the present invention.

1...signal processing system, 11...signal processing device, 21, 21-1 to 21-N...measurement sensors, 22-1 to 22-M...reference sensors, 31...measurement signal acquisition unit, 32...reference signal acquisition unit, 33...signal processing unit, 34...output unit, 41...noise component extraction unit, 42...target signal extraction unit, 1011, 1211...measurement signal, 1021, 1231...target signal, 1022...noise component, 1031, 1221, 2011 to 2015...reference signal, 2021 to 2026...noise candidate component, 2031...characteristics

Claims (11)

The computer
A signal processing method for extracting, as a noise component, a common component between a measurement signal measured in a state where a target signal and a noise signal are mixed, and a reference signal measured in the presence of the noise signal, the method comprising:
The computer,
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components corresponding to singular values exceeding a predetermined threshold value or one or more components corresponding to singular values equal to or greater than the predetermined threshold value from among a plurality of components that are candidates for the noise components, and extracting the selected components as the noise components;
Signal processing methods. The computer
A signal processing method for extracting, as a noise component, a common component between a measurement signal measured in a state where a target signal and a noise signal are mixed, and a reference signal measured in the presence of the noise signal, the method comprising:
The computer,
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components selected by a user based on the content of an operation performed by the user from among a plurality of components that are candidates for the noise component, and extracting the selected components as the noise component;
Signal processing methods. The computer,
calculating a correlation between the extracted noise component and the measurement signal, and based on the result of the correlation, correcting the noise component by multiplying the amplitude or intensity of the extracted noise component by a predetermined number, using a predetermined number corresponding to a value that brings the noise component closer to the noise component contained in the measurement signal while maintaining the waveform of the noise component , and removing the corrected noise component from the measurement signal;
3. The signal processing method according to claim 1. The measurement signal and the reference signal are measured during a common time period,
The computer,
extracting the common component as the noise component by performing signal processing by corresponding the signal portions measured at the same time to each other ;
The signal processing method according to any one of claims 1 to 3 . the measurement signal and the noise signal are signals measured during the same time period from a same measurement start timing to a same measurement end timing,
the same time period corresponds to the common time period,
5. The signal processing method according to claim 4 . The computer,
When there are a plurality of the measurement signals, processing is performed for each of the measurement signals.
The signal processing method according to any one of claims 1 to 5 . the measurement signal and the reference signal are magnetic signals;
The signal processing method according to any one of claims 1 to 6 . a measurement signal acquiring unit that acquires a measurement signal measured in a state in which a target signal and a noise signal are mixed;
a reference signal acquisition unit that acquires a reference signal measured in the presence of the noise signal;
a signal processing unit that extracts a common component between the measurement signal acquired by the measurement signal acquisition unit and the reference signal acquired by the reference signal acquisition unit as a noise component;
Equipped with
The signal processing unit includes:
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components corresponding to singular values exceeding a predetermined threshold value or one or more components corresponding to singular values equal to or greater than the predetermined threshold value from among a plurality of components that are candidates for the noise components, and extracting the selected components as the noise components;
Signal processing device. a measurement signal acquiring unit that acquires a measurement signal measured in a state in which a target signal and a noise signal are mixed;
a reference signal acquisition unit that acquires a reference signal measured in the presence of the noise signal;
a signal processing unit that extracts a common component between the measurement signal acquired by the measurement signal acquisition unit and the reference signal acquired by the reference signal acquisition unit as a noise component;
Equipped with
The signal processing unit includes:
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components selected by a user based on the content of an operation performed by the user from among a plurality of components that are candidates for the noise component, and extracting the selected components as the noise component;
Signal processing device. On the computer,
A first step of acquiring a measurement signal measured in a state where a target signal and a noise signal are mixed;
a second step of obtaining a reference signal measured in the presence of the noise signal;
a third step of extracting a common component between the acquired measurement signal and the acquired reference signal as a noise component;
A program for executing
The third step is
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components corresponding to singular values exceeding a predetermined threshold value or one or more components corresponding to singular values equal to or greater than the predetermined threshold value from among a plurality of components that are candidates for the noise components, and extracting the selected components as the noise components;
program. On the computer,
A first step of acquiring a measurement signal measured in a state where a target signal and a noise signal are mixed;
a second step of obtaining a reference signal measured in the presence of the noise signal;
a third step of extracting a common component between the acquired measurement signal and the acquired reference signal as a noise component;
A program for executing
The third step is
Using a singular value decomposition technique, vectors constituting the measurement signal and the reference signal are extracted, and singular values that are values representing the magnitude of the influence that the vectors have on constituting the signals are detected;
selecting one or more components selected by a user based on the content of an operation performed by the user from among a plurality of components that are candidates for the noise component, and extracting the selected components as the noise component;
program.

Images