JP7593000B2 - SIGNAL SEPARATION DEVICE, PROGRAM, AND SIGNAL SEPARATION METHOD - Google Patents

Description

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

Conventionally, interference signals and artifacts, which are noises that are simultaneously measured in biomagnetic measurement devices such as magnetoencephalography (MEG) measuring devices, need to be removed because they impede analysis.

Patent document 1 discloses a technology for separating and removing signals into signal subspaces in order to remove interference signals or artifacts in multi-channel magnetic field or charge potential measurements, and for spatially separating signals using a physical model that uses the Maskwell equation.

However, conventional techniques for removing interference signals and artifacts using spatial separation require setting parameters, but in the past there was no indicator of whether the parameters were appropriate, and trials had to be performed using the final processing results, so there was a demand for a method to more easily perform appropriate spatial separation.

The present invention has been made in consideration of the above, and aims to create an appropriate subspace for a target signal.

In order to solve the above-mentioned problems and achieve the object, the present invention provides a signal separation device that separates signals using spatial separation, wherein, when separating a target signal from signals other than the target signal from spatial information of a plurality of sensors installed at different positions, the device is configured to separate the signals according to spatial information on the plurality of sensors installed at different positions, and is configured to calculate a degree of agreement between a first region and a second region, the first region being a partial spatial region of the target signal and the second region being a location where the target signal is generated, a spatial separation parameter is set based on the degree of agreement, a degree of separation is determined as an index showing whether the set spatial separation parameter is appropriate, and the target signal and signals other than the target signal are separated from each other by the spatial separation parameter set using the degree of separation, and the partial space of the target signal is a vector space in which the projected spatial information can be separated into signals including the target signal and signals not including the target signal.

The present invention has the effect of creating an appropriate subspace for the target signal.

FIG. 1 is a diagram illustrating an example of the configuration of an information processing system according to the first embodiment. FIG. 2 is a block diagram showing an example of a hardware configuration of an information processing device. FIG. 3 is a block diagram showing an artifact removal function provided in the information processing apparatus. FIG. 4 is a flowchart showing the flow of the parameter determination process. FIG. 5 is a diagram showing an example of a region of interest divided into voxels. FIG. 6 is a diagram for explaining the spread of the subspace of the target signal and its index (DSSP method). FIG. 7 is a diagram illustrating the spread of the subspace of the target signal and its index (tSSS method) according to the second embodiment. FIG. 8 is a flowchart showing a flow of creating a subspace of a target signal according to the third embodiment. FIG. 9 is a diagram for explaining the subspace index (DSSP method) of the target signal.

Embodiments of a signal separation device, a program, and a signal separation method are described in detail below with reference to the attached drawings.

(First embodiment)
<System Configuration>
FIG. 1 is a diagram showing an example of the configuration of an information processing system 1 according to the first embodiment. In FIG. 1, the information processing system 1 measures and displays a plurality of types of biosignals, such as magnetoencephalography (MEG) signals and electroencephalography (EEG) signals. The information processing system 1 includes a measurement device 3, a data recording server 42, and an information processing device 20. The information processing device 20 has a monitor display 26 that displays signal information (measurement information) obtained by measurement and analysis results. Here, the data recording server 42 and the information processing device 20 are depicted separately, but at least a part of the data recording server 42 may be incorporated into the information processing device 20.

The subject lies on his/her back on the measurement table 4 with electrodes (or sensors) for measuring EEG attached to his/her head, and places his/her head in the recess 31 of the Dewar 30 of the measurement device 3. The Dewar 30 is a holding container in an extremely low temperature environment using liquid helium, and inside the recess 31 of the Dewar 30, a number of magnetic sensors for measuring magnetoencephalography are arranged. The measurement device 3 collects EEG signals from the electrodes and magnetoencephalography signals from the magnetic sensors, and outputs the collected biosignals (measurement information) to the data recording server 42. The measurement information recorded in the data recording server 42 is read out by the information processing device 20, where it is displayed and analyzed. Generally, the Dewar 30 containing the magnetic sensors and the measurement table 4 are arranged in a magnetically shielded room, but for convenience of illustration, the magnetically shielded room is omitted.

As described above, the measurement device 3 is a biomagnetic measurement device such as a magnetoencephalogram (MEG) measurement device, and functions as a multi-channel measurement device. The measurement device 3, which is a multi-channel measurement device, has known positions and orientations of multiple magnetic sensors used for magnetoencephalography, and can calculate the response values of the magnetic sensors when a signal source close to the magnetic sensor group is assumed.

The information processing device 20 displays the waveforms of magnetoencephalographic signals from multiple magnetic sensors and the waveforms of electroencephalographic signals from multiple electrodes, synchronized on the same time axis. The electroencephalographic signals represent the electrical activity of nerve cells (the flow of ionic charges that occurs in the dendrites of neurons during synaptic transmission) as a voltage value between the electrodes. The magnetoencephalographic signals represent minute magnetic field fluctuations caused by electrical activity in the brain. The brain magnetic field is detected by a highly sensitive superconducting quantum interference device (SQUID) sensor.

<Hardware Configuration>
2 is a block diagram showing an example of a hardware configuration of the information processing device 20. The information processing device 20 has a CPU (Central Processing Unit: processor) 21, a RAM (Random Access Memory) 22, a ROM (Read Only Memory) 23, an auxiliary storage device 24, an input/output interface 25, and a monitor display (display device) 26. The CPU 21, the RAM 22, the ROM 23, the auxiliary storage device 24, the input/output interface 25, and the monitor display (display device) 26 are connected to each other via a bus 27.

The CPU 21 controls the overall operation of the information processing device 20 and performs various information processing. The CPU 21 also executes information processing programs stored in the ROM 23 or the auxiliary storage device 24 to control the display operation of the measurement recording screen and the analysis screen.

The information processing program executed by the information processing device 20 of this embodiment is provided in the form of an installable or executable file recorded on a computer-readable recording medium such as a CD-ROM, a flexible disk (FD), a CD-R, or a DVD (Digital Versatile Disk).

The information processing program executed by the information processing device 20 of this embodiment may be stored on a computer connected to a network such as the Internet and provided by downloading it via the network. The information processing program executed by the information processing device 20 of this embodiment may be provided or distributed via a network such as the Internet.

In addition, the information processing program executed by the information processing device 20 of this embodiment may be configured to be provided by being pre-installed in a ROM or the like.

RAM 22 is used as a work area for CPU 21 and may include a non-volatile RAM for storing main control parameters and information. ROM 23 stores basic input/output programs, etc. The information processing program of the present invention may also be stored in ROM 23.

The auxiliary storage device 24 is a storage device such as an SSD (Solid State Drive) or HDD (Hard Disk Drive), and stores, for example, information processing programs that control the operation of the information processing device 20, and various data and files necessary for the operation of the information processing device 20.

The input/output interface 25 includes both a user interface such as a touch panel, keyboard, display screen, and operation buttons, and a communication interface that inputs information from various sensors or the data recording server 42 and outputs analysis information to other electronic devices.

The monitor display 26 displays a measurement recording screen and an analysis screen, and the screen is updated according to input/output operations via the input/output interface 25.

<Functional configuration>
Next, a description will be given of the function of the information processing device 20 according to this embodiment, which is a function of removing interference signals and artifacts, which are noises simultaneously measured by the measurement device 3. Fig. 3 is a block diagram showing the artifact removal function provided in the information processing device 20.

The information processing device 20 functions as a signal separation device and includes a measurement information acquisition means 201, a parameter setting means 202, a spatial separation means 203, a time separation means 204, and an artifact separation means 205.

The measurement information acquisition means 201, the parameter setting means 202, the spatial separation means 203, the time separation means 204, and the artifact separation means 205 are realized by the CPU 21 reading and executing an information processing program stored in the ROM 23 or the auxiliary storage device 2.

The measurement information acquisition means 201 acquires the biosignal (measurement information) measured by the measurement device 3 from the data recording server 42. The measurement information acquisition means 201 stores the biosignal (measurement information), which is sensor information measured by the measurement device 3, in the auxiliary storage device 24.

The parameter setting means 202 uses spatial information, which is the position and orientation of the magnetic sensor of the measuring device 3, to select and set appropriate parameters for the spatial separation described below.

That is, the parameter setting means 202 sets the value calculated based on the extent of the subspace of the target signal as the parameter.

The parameter setting means 202 may select a parameter based on a predetermined criterion related to the extent of the subspace of the target signal from the parameter options and set it as the parameter.

The parameter setting means 202 may also set a value calculated based on the extent of the subspace of the target signal as a parameter by replacing a preset value.

The spatial separation means 203 separates the target signal into an inside and outside subspace using a projection matrix created based on the parameters set by the parameter setting means 202.

The time separation means 204 finds the basis vectors in the time direction of each subspace separated by the space separation means 203 using singular value decomposition or the like, and separates out components other than the target signal.

The artifact separation means 205 functions as a signal separation means, and removes interference signals and artifact components based on basis vectors in the time direction. The artifact separation means 205 also determines the degree of (spatial) separation (in this embodiment, the degree of separation is the separation gain) as an index of whether the parameters set in the parameter setting means 202 are appropriate.

Here, we explain an example of how parameters are determined in the measurement device 3.

Figure 4 is a flowchart showing the flow of the parameter determination process. Here, we will describe an example of processing using the DSSP method (see Non-Patent Document 1), which is one of the methods for removing interference signals and artifacts using spatial separation.

As shown in Fig. 5, the spatial separation means 203 divides a preset region of interest (a region where a target signal to be observed may occur) into voxels V. Then, as shown in Fig. 4, the spatial separation means 203 calculates, for all voxels V in the region of interest, a sensor response L(r i ) in the case where a signal source is present in the r _i -th voxel V in the region of interest, from spatial information on the positions and orientations of the magnetic sensors of the Dewar 30 of the measuring device 3 equipped with _M magnetic sensors, and calculates prior information F as shown in the following formula (1) (step S11).

Next, the time separation means 204 calculates a projection matrix P onto the subspace of the target signal by using the eigenvectors of the Gram matrix of the prior information F calculated in step S11 as basis vectors of the subspace of the target signal, as shown in the following formulas (2) and (3) (step S12). When the time separation means 204 calculates the projection matrix P onto the subspace of the target signal, the projection matrix is created by compressing the dimensions of the prior information, and this becomes the dimension ζ (0<ζ<M) of the subspace of the target signal.

The above processing is known as part of the DSSP method, which is a method for removing interference signals and artifacts using spatial separation.

Next, the artifact separation means 205 calculates a quantitative index of the subspace of the target signal to evaluate the performance of the projection of the projection matrix P, which projects the responses B(x, y, z) of the multiple magnetic sensors of the measuring device 3 in a signal generated from an arbitrary position onto the subspace of the target signal and separates the signals into signal 1 that includes the target signal and signal 2 that does not (step S13).

More specifically, the artifact separation means 205 defines a quantitative index of the subspace of the target signal using the intensity ratio of the original signal B(x, y, z) to the separated signal, as shown in the following formula (4). Note that in formula (4), signal 2 ((I-P)B) that does not contain the target signal is used, but signal 1 (PB) that contains the target signal is used as well, and what is important here is the proportion of the signal generated from that position that is separated into signal 1 and signal 2. Here, the case of projection onto the outer subspace will be explained. The artifact separation means 205 calculates the subspace gain φ(x, y, z) for B(x, y, z) at various positions.

Then, the artifact separating means 205 calculates the degree of coincidence between the region (white region) R _φ where the gain φ is equal to or less than the threshold value and the region of interest R _I set in step S11 (step S14). In this embodiment, other indices are also added, and the score λ of the Jaccard Index, which is one of the indices of the degree of coincidence, is calculated as shown in the following formula (5).

Other methods for calculating similarity include general similarity measures such as Dice Index and volume similarity, evaluation indices for binary classification models such as Accuracy and F-value, and similarity measures using probability.

Even if binarization is not performed, it can be obtained using various probability distributions and their evaluation indices, such as ROC curves and their Area Under the Curve, and image comparison methods. It is also possible to optimize the above-mentioned threshold value by using probability distributions.

Next, the artifact separation means 205 determines ζ that creates the subspace of the target signal that is most suitable for the region of interest by finding ζ that maximizes the score λ as shown in equation (6) below (step S15).

The above processing is effective for methods that have parameters that affect the spread of the subspace of the target signal, similar to the DSSP method.

Next, we will explain the extent of the subspace of a specific target signal and its index (DSSP method).

Here, Figure 6 is a diagram explaining the spread of the subspace of the target signal and its index (DSSP method). (a), (b), and (c) in Figure 6 are diagrams that visualize the value of the gain φ by changing the value of the dimension ζ of the subspace of the target signal.

As shown in FIG. 6, the spatial separation means 203 sets the region of interest (the region where the signal to be kept occurs) as a rectangular parallelepiped divided into voxels V. Then, the spatial separation means 203 cuts out and displays a specific axial cross section of the set rectangular parallelepiped. In other words, the white frames (a), (b), and (c) in FIG. 6 represent the range of the set region of interest. In (a), (b), and (c) in FIG. 6, the stronger the white color, the smaller the gain for separating the signal generated from that position into signal 2 that does not contain the target signal, that is, it means that most of the signal generated from that position is separated into signal 1 that contains the target signal. Therefore, based on the premise that the signal within the region of interest is the target signal, it is ideal that the white color is stronger in the white frame and suppressed and black in other areas. Therefore, the white area represents the spread of the subspace of the target signal.

Figure 6(a) shows that the subspace of the target signal is small and insufficient relative to the defined region of interest. On the other hand, Figure 6(c) shows that the subspace of the target signal is large relative to the defined region of interest, and interference signals outside the region of interest are also incorporated into the target signal. Therefore, while it is clear that Figure 6(b) is optimal, it is also possible to select optimal parameters by quantifying the spread by determining the degree of coincidence between the region of interest and the subspace of the target signal.

Figure 6 (A), (B), and (C) show the waveforms resulting from removing the interference signal using the above-mentioned Figures 6 (a), (b), and (c), respectively. The interference signal used in this case was an interference signal recorded by measuring device 3, a magnetoencephalography (MEG) measuring device, with a noise source actually placed near the left clavicle. Figures 6 (A), (B), and (C) show data in which the simulated biosignal of Figure 6 (D) has been superimposed on this interference signal. It is clear from the naked eye that the waveform of Figure 6 (B) is closest to the correct waveform of Figure 6 (D).

For this reason, the final result will be better if the extent of the subspace of the target signal is matched to the region of interest, so the artifact separation means 205 determines parameters using the extent of the subspace of the target signal as an index.

In this way, according to this embodiment, by calculating the separation gain as an index of whether the set parameters are appropriate, it is possible to create an appropriate subspace for the target signal.

Therefore, according to this embodiment, it is not necessary to search for parameters by trial and error after measurement. Furthermore, according to this embodiment, the results do not change due to different parameters selected by different analysts. Furthermore, according to this embodiment, in the case of magnetoencephalography, a subspace of the target signal tailored to each patient can be created.

Second Embodiment
Next, a second embodiment will be described.

The second embodiment differs from the first embodiment, which uses the DSSP method, in that it uses the tSSS method. In the following explanation of the second embodiment, the explanation of the same parts as in the first embodiment will be omitted, and only the differences from the first embodiment will be explained.

Figure 7 is a diagram explaining the spread of the subspace of the target signal and its index (tSSS method) in the second embodiment.

In the tSSS method, the center of the sphere is a parameter related to the extent of the subspace of the target signal, so (a), (b), and (c) in Figure 7 visualize the value of the gain φ by changing the center of the sphere.

Note that because the way the subspace of the target signal spreads is different from that of the DSSP method shown in Figure 6, Figure 7 shows the sagittal central cross section, with the right side of Figure 7 showing the anterior and the left side showing the posterior. Also, because the region of interest in the tSSS method is a sphere, Figure 7 also projects the position of the sensor assumed to be installed in this case.

The region of interest shown in Figure 7 is the inside of the area where the sensor is installed. Note that in (a), (b), and (c) shown in Figure 7, the stronger the white, the smaller the gain in separating the signal generated from that position into signal 2 that does not contain the target signal, meaning that most of the signal generated from that position is separated into signal 1 that does contain the target signal. Therefore, based on the premise that the signal within the region of interest is the target signal, ideally the region at the center of the sphere will have a stronger white color, while other areas will be suppressed and turn black. Thus, the white region represents the spread of the subspace of the target signal.

Figure 7(a) shows that the subspace of the target signal is small and insufficient relative to the defined region of interest, while Figure 7(c) shows that the subspace of the target signal is large relative to the defined region of interest, and interference signals outside the region of interest are also incorporated into the target signal. Therefore, while it can be seen that Figure 7(b) is optimal, it is also possible to select optimal parameters by quantifying the spread by determining the degree of coincidence between the region of interest and the subspace of the target signal.

In addition, (A), (B), and (C) in Figure 7 are the waveforms resulting from removing the interference signal using the above-mentioned Figures 7(a), (b), and (c), respectively. The interference signal used in this case was an interference signal recorded by measuring device 3, which is a magnetoencephalograph, with a noise source actually placed near the left clavicle. Figures 7(A), (B), and (C) are data in which the simulated biosignal in Figure 7(D) is superimposed on this interference signal. It is clear from the naked eye that the waveform in Figure 7(B) is closest to the correct waveform in Figure 7(D).

For this reason, the final result will be better if the extent of the subspace of the target signal is matched to the region of interest, so the artifact separation means 205 determines parameters using the extent of the subspace of the target signal as an index.

In this manner, according to this embodiment, the tSSS method can also quantify the spread of the subspace of the target signal from the ratio calculated for any signal source, and appropriate parameters can be determined based on the spread of the quantified subspace of the target signal, thereby creating an appropriate subspace of the target signal.

Third Embodiment
Next, a third embodiment will be described.

The third embodiment describes a method for creating a subspace of a target signal in the measurement device 3. In the following description of the third embodiment, the same parts as in the first embodiment will be omitted, and only the differences from the first embodiment will be described.

Here, FIG. 8 is a flowchart showing the flow of creating a subspace of a target signal according to the third embodiment. The flowchart shown in FIG. 8 details the processing by the spatial separation means 203 in step S11 of FIG. 4 described in the first embodiment.

As shown in FIG. 8, the spatial separation means 203 manually or automatically extracts the signal generation region using information obtained from an information acquisition device such as an imaging device or a shape measurement device (step S111).

For example, when the measuring device 3 is a magnetoencephalogram (MEG), the spatial separation means 203 aligns information obtained from an imaging device such as an MRI (Magnetic Resonance Imaging) with the magnetic sensor of the Dewar 30 of the measuring device 3, thereby creating voxels or meshes to match the region of interest with the actual signal generation region. Here, the signal generation region refers to, for example, the brain region in the case of a magnetoencephalogram (MEG), and refers to the region where the target signal to be observed can be generated.

Next, the spatial separation means 203 calculates the prior information F within the signal generation area using the sensor information measured by the measurement device 3 acquired via the measurement information acquisition means 201 (step S112).

Here, FIG. 9 is a diagram explaining the subspace index (DSSP method) of the target signal. FIG. 9(a) is a waveform resulting from executing DSSP using a projection matrix created with voxels defined in step S111 of FIG. 8. On the other hand, FIG. 9(b) is a waveform resulting from executing DSSP using a projection matrix created with voxels defined in a rectangular parallelepiped as described in FIG. 6. The interference signal used in this case is an interference signal recorded by measuring device 3, which is a magnetoencephalography (MEG) measuring device, with a noise source actually placed near the left clavicle. FIG. 9(a) and FIG. 9(b) are data in which the simulated biological signal of FIG. 9(c) is superimposed on this interference signal. It is clear that the waveform of FIG. 9(a) is closest to the waveform of FIG. 9(c).

This shows that when creating voxels, it is useful to use a method of creating voxels by matching the region of interest to the actual shape of the brain, based on information obtained from imaging equipment such as MRI and shape measurement devices, and information obtained by aligning the magnetic sensor of the Dewar 30 of the measurement device 3.

Methods for representing the actual brain shape include, but are not limited to, approximating it with a single sphere (single-sphere model), approximating it with multiple spheres (overlapping-sphere model), or extracting the actual brain shape from imaged information.

In this way, according to this embodiment, by creating a space by setting voxels that correspond to the signal generation area (the area where the target signal to be observed may occur), spatial separation performance is improved, and interference signal removal performance can be further improved when appropriate parameters are selected.

In each embodiment, the living body is described in detail by taking the head as an example, but the living body is not limited to this. For example, the living body may be the cervical vertebrae that are the subject of a magnetospinography (MagnetoSpinoGraphy (MSG)), and the tracking of the living body in a magnetospinography measurement is also within the scope of the present invention.

20 Signal separation device 202 Parameter setting means 203 Spatial separation means 204 Time separation means 205 Signal separation means

Patent No. 4875696 K. Sekihara, Y. Kawabata, S. Ushio, S. Sumiya, S. Kawabata, Y. Adachi, S. S. Nagarajan, “Dual signal subspace projection (DSSP): a novel algorithm for removing large interference in biomagnetic measurements,” Journal of Neural Engineering, vol.13, no.3, 036007, 2016

Claims (9)

1. A signal separation device that separates signals using spatial separation,
When separating a target signal from signals other than the target signal from spatial information of a plurality of sensors installed at different positions, the separation is performed according to spatial information on the plurality of sensors installed at different positions;
configured to calculate a degree of match between a first region and a second region, the first region being a subspace region of the target signal and the second region being where the target signal is generated;
setting a spatial separation parameter based on the degree of match;
A degree of separation is determined as an index showing whether the set spatial separation parameter is appropriate, and a target signal and a signal other than the target signal are separated from each other by the spatial separation parameter set using the degree of separation,
The subspace of the target signal is a vector space in which the projected spatial information can be separated into a signal including the target signal and a signal not including the target signal.
A signal separation device comprising: a spatial separation means for setting a region in which the target signal may occur, and separating the target signal into an inside and an outside of a subspace region using a projection matrix created based on the set spatial separation parameters;
2. The signal separating device according to claim 1. A time separation means is provided for calculating a basis vector in a time direction of each of the subspace regions separated by the space separation means, and separating components other than the target signal.
3. The signal separating device according to claim 2. 1. A signal separation device that separates signals using spatial separation,
a spatial separation means for separating signals using the projection matrix created when creating a projection matrix for separating a target signal from signals other than the target signal from spatial information of a plurality of sensors installed at different positions, the spatial separation means setting a space corresponding to an area where the target signal may occur so as to coincide with the area where the target signal may occur, depending on a degree of separation of the target signal by the projection matrix ;
A signal separation device comprising: the spatial separation means sets a space corresponding to a region where the target signal may be generated, using information about the region of interest obtained from an imaging device or a shape measurement device;
5. The signal separating device according to claim 4. A computer that controls a signal separation device that separates signals using spatial separation,
a parameter setting means configured to separate a target signal from a signal other than the target signal according to spatial information on a plurality of sensors installed at different positions, when separating the target signal from the spatial information on the plurality of sensors installed at different positions, and configured to calculate a degree of agreement between a first region and a second region, the first region being a partial spatial region of the target signal and the second region being a location where the target signal is generated, and to set a spatial separation parameter based on the degree of agreement;
a signal separating means for determining a degree of separation as an index showing whether the spatial separation parameters set by the parameter setting means are appropriate, and for separating a target signal from a signal other than the target signal by the spatial separation parameters set using the degree of separation;
Function as a
The subspace of the target signal is a vector space in which the projected spatial information can be separated into a signal including the target signal and a signal not including the target signal.
Program for. A computer that controls a signal separation device that separates signals using spatial separation,
a spatial separation means for separating a target signal from signals other than the target signal based on spatial information from a plurality of sensors installed at different positions, the spatial separation means setting a space corresponding to an area where the target signal may be generated so as to coincide with the area where the target signal may be generated, depending on a degree of separation of the target signal by the projection matrix, and separating signals using the projection matrix thus generated ;
A program to function as a A signal separation method in a signal separation device that separates signals using spatial separation, comprising:
a parameter setting step of setting a spatial separation parameter based on the degree of matching between a first region and a second region, the parameter setting step being configured to perform separation according to spatial information on a plurality of sensors installed at different positions when separating a target signal from a signal other than the target signal from spatial information on the plurality of sensors installed at different positions, the parameter setting step being configured to calculate a degree of matching between a first region and a second region, the first region being a partial spatial region of the target signal and the second region being a location where the target signal is generated;
a signal separation step of determining a degree of separation as an index showing whether the spatial separation parameters set in the parameter setting step are appropriate, and separating a target signal from a signal other than the target signal using the spatial separation parameters set using the degree of separation;
Including,
The subspace of the target signal is a vector space in which the projected spatial information can be separated into a signal including the target signal and a signal not including the target signal.
2. A signal separation method comprising: A signal separation method in a signal separation device that separates signals using spatial separation, comprising:
A spatial separation step of creating a projection matrix for separating a target signal from signals other than the target signal from spatial information of a plurality of sensors installed at different positions, the spatial separation step being to set a space corresponding to an area where the target signal may be generated so as to coincide with the area where the target signal may be generated , depending on a degree of separation of the target signal by the projection matrix, and to separate signals using the created projection matrix .
2. A signal separation method comprising:

Images