JP2006178314A - Device for separating and extracting mixed signal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device for separating/extracting a mixed signal capable of reproducing an original signal continuously and faithfully. <P>SOLUTION: The device for separating/extracting a mixed signal obtains signal separation outputs XA (XA<SB>1</SB>, ..., XA<SB>i</SB>) and XB (XB<SB>1</SB>, ..., XB<SB>i</SB>) by performing separation/extraction processing by an independent component analysis of each two continuous sampling periods Δt1, Δt2. Further, the device produces a signal separation output XC (XC<SB>1</SB>, ..., XC<SB>i</SB>) in Δt1+Δt2, and divides XC into a component (XA') corresponding to the period Δt1 and a component (XB') corresponding to Δt2. Then, the device calculates combinations to couple each component of (XA<SB>1</SB>, ..., XA<SB>i</SB>) to any of the components (XB<SB>1</SB>, ..., XB<SB>i</SB>) based on a correlation value between XA and XA', and that between XB and XB'. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a mixed signal separation / extraction apparatus. In particular, the present invention relates to an improvement in a technique for separating and extracting an original signal from mixed signals individually observed by a plurality of sensors.

  In order to separate and extract each original signal from a mixed signal including a plurality of original signals, a blind signal separation method (BBS: Blind Source Separation) is used (see, for example, Patent Document 1). By combining independent component analysis with the blind signal separation method, a plurality of signals can be separated without prior knowledge of the incoming signal. If N sensors that receive the mixed signal are used, a maximum of N signal components can be separated. In the blind signal separation method, digital data obtained by sampling a mixed signal reaching each sensor at a predetermined time interval is used.

  By the way, in order to separate and extract the original signal by independent component analysis, it is necessary to rely only on the independence of the received mixed signal. For this reason, the separated / extracted signals remain in the wrong order for each sampling period and the arbitraryness of the size. When the period for separation / extraction is determined, such a phenomenon does not affect the waveform itself, so that there is no serious problem in terms of signal separation / extraction.

However, when the separation / extraction process is continuously performed, or when the number and state of the original signals change every moment, it is necessary to output the reception signal separation / extraction results at regular time intervals. If the arbitrary separation of the separated and extracted signals occurs in such processing, the continuity of the signals is impaired.
JP 2003-92557 A

As described above, in the original signal separated and extracted by the blind signal separation method, the difference in order for each sampling period and the arbitraryness of the size remain. Therefore, when the separation / extraction process is performed continuously (that is, at regular time intervals), the original order may be lost and the original signal may not be output as a continuous signal.
The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a mixed signal separating / extracting apparatus capable of reproducing an original signal continuously and faithfully.

  In order to achieve the above object, according to an aspect of the present invention, a plurality of sensors each receiving a mixed signal including a plurality of original signals, and the original signal is analyzed by independent component analysis from the received signals for each of the plurality of sensors. In the mixed signal separating / extracting device including the independent component analyzing unit for separating / extracting, the independent component analyzing unit acquires the sample data by sampling the received signals for each of the plurality of sensors for each continuous sampling period. Sampling means, separation means for generating a plurality of original signal data for each sampling period by independent component analysis using the sample data, and a plurality generated in the preceding first sampling period of two consecutive sampling periods In the second sampling period following the first sampling period. A calculation means for calculating a combination to be connected to the plurality of second original signal data, and a combination of the plurality of first original signal data calculated by the calculation means. There is provided a mixed signal separating / extracting device, characterized in that the mixed signal separating / extracting device is provided with connecting means for reproducing a continuous original signal by connecting to any of the original signal data.

  By taking such means, the original signals separated and extracted in the sampling periods that are temporally continuous are combined and connected based on, for example, the degree of correlation strength. That is, it is possible to eliminate the difference in the order of each sampling period and the arbitraryness of the magnitude based on the characteristics of each original signal, and it is possible to accurately connect consecutive original signals. Therefore, it is possible to provide a mixed signal separating / extracting apparatus capable of reproducing the original signal continuously and faithfully.

  ADVANTAGE OF THE INVENTION According to this invention, the mixed signal separation / extraction apparatus which can reproduce | regenerate an original signal continuously and faithfully can be provided.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The mixed signal separating / extracting apparatus according to this embodiment is a method for blind signal separation of a plurality of original signals (m) simultaneously arriving at a plurality of sensors (n) arranged in an array. To separate and extract. Blind signal separation is an algorithm that separates up to the same number of incident signals as the number of sensors without prior knowledge of sensor responsiveness and signal characteristics and incident signals. This algorithm is called blind signal separation or independent component analysis (ICA), and many known documents have been published.

[First Embodiment]
The mixed signal separating / extracting apparatus according to the first embodiment of the present invention uses an antenna as a sensor, and separates and outputs an original signal from a mixed signal of a plurality of incident signals incident on a plurality of antennas. In the following, for simplicity of explanation, the number of antennas and the number of incident signals are both assumed to be four, but the number of antennas and the number of incident signals are not limited to these and are arbitrary.

  FIG. 1 is a functional block diagram showing an embodiment of a mixed signal separating / extracting apparatus according to the present invention. This mixed signal separating / extracting device includes a first antenna 11, a second antenna 12, a third antenna 13, a fourth antenna 14, a first band limited filtering device F1, a second band limited filtering device F2, a third band. A limiting filtering device F3, a fourth band limiting filtering device F4, a sampling processing unit 100, an independent component analysis unit 200, and an output processing unit 400 are provided.

  As the first to fourth antennas 11 to 14, a non-directional antenna such as a vertical antenna or a dipole antenna, an antenna having an arbitrary directivity, or the like is used, and receives radio waves from various directions. The interval and height at which these first to fourth antennas 11 to 14 are installed are arbitrary.

  The first antenna 11 receives a plurality of incident signals (radio waves) X1 to X4 from the air, and sends a mixed signal in which these signals are mixed to the first band-limited filtering device F1. Similarly, the second to fourth antennas 12 to 14 receive a plurality of incident signals X1 to X4 from the air, and the mixed signals obtained by mixing these signals are sent to the second to fourth band-limited filtering devices F2 to F4. Send each one.

  The first band-limited filtering device F <b> 1 passes only the frequency components in the predetermined band included in the mixed signal from the first antenna 11, and sends them to the sampling processing unit 100. Similarly, the second to fourth band-limited filtering devices F2 to F4 pass only frequency components in a predetermined band included in the mixed signals from the second to fourth antennas 12 to 14 to the sampling processing unit 100, respectively. send. In addition, the frequency band which each 1st-4th band limitation filter apparatus F1-F4 passes is the same.

  FIG. 2 is a schematic diagram schematically showing the independent component analysis unit 200 of FIG. In FIG. 2, the sampling processing unit 100 samples the outputs of the first to fourth band-limited filtering devices F1 to F4 for each successive sampling period, and obtains input signals Xa,. The input signals Xa,..., Xj are acquired in the sampling periods Δt1, Δt2,. These input signals Xa,..., Xj are input to the mixed signal connection processing unit 300 and connected to output a continuous signal.

The following expression (1) is an expression expressing the independent component analysis method as a model. In the present embodiment, S _{1,..., M} (t) in Expression (1) is calculated by a blind signal separation method.

By the way, in general, there is an arbitrary order of randomness between the signals X1 _{, ..., n} (t) separated by the blind signal separation and the original signals S1 _{, ..., m} (t). In the following, a technique for dealing with this arbitraryness and concatenating the separated signals in an appropriate order will be described. The sampling interval of the sampling processing unit 100 is a value that is not affected by the time difference of arrival of the mixed signal that reaches each sensor.

FIG. 3 is a functional block diagram showing in more detail the first embodiment of the independent component analysis unit 200 of FIG. In FIG. 3, the input signal is sampled every two sampling periods Δt1 and Δt2 that are continuous in time, and input signals Xa (Xa ₁ ,..., Xa _i ) and Xb (Xb ₁ ,..., Xb _i ) are obtained, respectively. It is done. Separation / extraction processing is performed for each input signal, and signal separation outputs XA (XA ₁ ,..., XA _i ), XB (XB ₁ ,..., XB _i ) are obtained.

In the present embodiment, signal separation outputs XC (XC ₁ ,..., XC _i ) are generated in a period corresponding to Δt1 + Δt2, that is, a period in which two sampling periods are connected. XC is divided into a component (XA ′) corresponding to the period Δt1 and a component (XB ′) corresponding to Δt2. A combination to be connected to each component of (XA ₁ ,..., XA _i ) and each component of (XB ₁ ,..., XB _i ) is a correlation value between XA and XA ′, and XB and XB ′. It is calculated based on the correlation value.

  When the combination is calculated, the terminal selection switch 500 is controlled based on the result, and terminals to which signals having a large cross-correlation value are input are connected to each other. As a result, XA and XB are connected in an appropriate combination. Therefore, it is possible to output continuous signals that are separated and extracted at the times of Δt1 and Δt2 and independent of each other.

  As the sampling period becomes longer, there may be some correlation between the incident signals. When separation / extraction is performed by blind signal separation in this state, separation between signals may not be sufficient, and other incident signal components may be mixed without being separated. This becomes a problem when a long-time signal or a continuous signal is separated and extracted. In general, the shorter the sampling period is, the smaller the correlation is, so that it is easy to separate and extract independent signals.

  Therefore, in the present embodiment, as shown in FIG. 2, a signal with a constant interval is divided into short arbitrary time interval signals Δt1, Δt2,..., Δti, and as a result of frequency analysis with the signal with a constant interval, a correlation value or a high value is obtained. The mixed signal linking means calculated using the next statistic eliminates the mistake in the order of the signals, and the continuous signals can be separated and extracted.

  That is, in the present embodiment, in order to connect the signals separated and extracted in two consecutive sampling periods, the results of separation and extraction in the combined period of each sampling period are used. Although the degree of separation of the signals separated and extracted for each sampling period is high, the arbitraryness of the order remains. On the contrary, the signals separated and extracted in the combined period of each sampling period have a low degree of separation but do not need to consider the arbitraryness of the order. Therefore, by calculating the correlation between the two and connecting them with a combination having a high correlation value, the original signal can be reproduced continuously and faithfully.

[Second Embodiment]
FIG. 4 is a functional block diagram showing the second embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. In FIG. 4, the frequency of the signal separation outputs XA (XA ₁ ,..., XA _i ) and XB (XB ₁ ,..., XB _i ) separated and extracted for each sampling period Δt 1 and Δt 2 is compared with each other to select a terminal. By selecting terminals having the same frequency in the switch 500, a combination of XA and XB to be connected is obtained. In this way, the original signal can be reproduced continuously and faithfully.

[Third Embodiment]
FIG. 5 is a functional block diagram illustrating the third embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. In FIG. 5, the frequency of the signal separation output (XA + XB) in the period (Δt1 + Δt2) and the signal separation outputs XA and XB separated and extracted for each of the sampling periods Δt1 and Δt2 are compared with each other. By selecting the terminals with the same X, the combination of XA and XB to be connected is obtained. In this way, the original signal can be reproduced continuously and faithfully.

[Fourth Embodiment]
FIG. 6 is a functional block diagram illustrating the fourth embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. In FIG. 6, the fourth-order statistics (fourth-order cumulants and fourth-order cumulants) of the signal separation output (XA + XB: XC) in the period (Δt1 + Δt2) and the signal separation outputs XA and XB separated and extracted for each of the sampling periods Δt1 and Δt2. Calculated). Based on the result, the terminal selection switch 500 selects a non-independent terminal to obtain a combination of XA and XB to be connected. In this way, the original signal can be reproduced continuously and faithfully.

[Fifth Embodiment]
FIG. 7 is a functional block diagram illustrating the fifth embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. In FIG. 7, the matrix switch 600 prepares all combinations of XA and XB in advance. Then, a correlation value between these sets and XCi is calculated, and the terminal selection switch 500 is controlled based on the calculated value to select a combination of XA and XB to be connected. In this way, the original signal can be reproduced continuously and faithfully.

[Sixth Embodiment]
FIG. 8 is a functional block diagram showing the sixth embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. Also in FIG. 8, all combinations of XA and XB are prepared in advance by the matrix switch 600. Then, a combination having a single frequency among these sets is selected by the terminal selection switch 500. In this way, the original signal can be reproduced continuously and faithfully.

[Seventh Embodiment]
FIG. 9 is a functional block diagram showing the seventh embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. Also in FIG. 9, the signal separation output XC in the period (Δt1 + Δt2) is obtained. Further, all combinations of XA and XB are prepared in advance by the matrix switch 600. Then, a quaternary statistic of these combinations and XC is calculated, and a terminal that is not independent is selected in the terminal selection switch 500 based on the result. Even in this way, the optimum solution of the combination of XA and XB can be obtained, and the original signal can be reproduced continuously and faithfully.

[Eighth Embodiment]
FIG. 10 is a functional block diagram showing the eighth embodiment of the independent component analyzer 200 of FIG. 1 in more detail. In FIG. 10, the signal separation outputs XAi and XBi are mixed (mixed or multiplied) over all combinations. All the mixed results are subjected to FFT (Fast Fourier Transform) processing and frequency analysis, and the spectrum of the baseband component and twice the harmonic component of XA (or XB) is detected.

  When XAi and XBi are mixed (mixed or multiplied), when the frequency components of XAi and XBi are the same, the spectrum of the doubled harmonic component is maximized. If the components are different, the value will be smaller. By utilizing this fact, an optimum solution of the combination of XA and XB is obtained by selecting the combination that maximizes the spectrum by the terminal selection switch 500. Even in this way, the original signal can be reproduced continuously and faithfully.

[Ninth Embodiment]
FIG. 11 is a functional block diagram illustrating the ninth embodiment of the independent component analysis unit 200 of FIG. 1 in more detail. In FIG. 11, all combinations of XA and XB are prepared in advance by the matrix switch 600, and the signal components of all combinations are connected. This concatenated signal and XC are mixed (mixed or multiplied) and subjected to FFT processing. Then, as in the eighth embodiment, an optimum solution for the combination of XA and XB is obtained by selecting the combination having the maximum spectrum by the terminal selection switch 500. Even in this way, the original signal can be reproduced continuously and faithfully.

[Tenth embodiment]
FIG. 12 is a functional block diagram showing in more detail the tenth embodiment of the independent component analysis unit 200 of FIG. In FIG. 12, the signal separation outputs XA and XB of the input signal Xa between Δt1 and the input signal Xb between Δt2 and Δt1 and Δt2 components XA ′ and XB ′ of the signal separation output XC at (Δt1 + Δt2) Mix (mix or multiply) each. The mixture component is subjected to FFT (frequency analysis) processing, and the combination of XA and XB is optimized by selecting the combination that maximizes the spectrum by the terminal selection switch 500, as in the eighth and ninth embodiments. Try to find a solution. Even in this way, the original signal can be reproduced continuously and faithfully.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the components without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

The functional block diagram which shows embodiment of the mixed signal separation / extraction apparatus concerning this invention. The schematic diagram which shows the independent component analysis part 200 of FIG. 1 roughly. The functional block diagram which shows 1st Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 2nd Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 3rd Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 4th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 5th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 6th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 7th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 8th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 9th Embodiment of the independent component analysis part 200 of FIG. 1 in detail. The functional block diagram which shows 10th Embodiment of the independent component analysis part 200 of FIG. 1 in detail.

Explanation of symbols

  F1 to F4: Band-limited filtering device, 11-14: Antenna, 100 ... Sampling processing unit, 200 ... Independent component analysis unit, 300 ... Mixed signal connection processing unit, 400 ... Output processing unit, 500 ... Terminal selection switch, 600 ... Matrix switch

Claims (10)

A mixed signal separation unit comprising: a plurality of sensors that respectively receive mixed signals including a plurality of original signals; and an independent component analysis unit that separates and extracts the original signals from the received signals of the plurality of sensors by independent component analysis. In the extraction device,
The independent component analyzer is
Sampling means for sampling the received signals for each of the plurality of sensors for each successive sampling period to obtain sample data;
Separation means for generating a plurality of original signal data for each sampling period by independent component analysis using the sample data;
A plurality of first original signal data generated in a preceding first sampling period of two consecutive sampling periods; a plurality of second original signal data generated in a second sampling period following the first sampling period; Calculating means for calculating a combination to be connected to each other;
Connecting means for connecting each of the plurality of first original signal data to any one of the plurality of second original signal data in a combination calculated by the calculating means, and reproducing a continuous original signal. A mixed signal separator / extractor. 2. The mixed signal separating / extracting apparatus according to claim 1, wherein the calculating unit calculates the combination based on a correlation between the plurality of first and second original signal data. 2. The mixed signal separating / extracting apparatus according to claim 1, wherein the calculating unit calculates the combination based on a frequency of the plurality of first and second original signal data. 2. The mixed signal separation / extraction according to claim 1, wherein the calculation unit calculates the combination based on a value of a fourth-order statistic calculated from the plurality of first and second original signal data. apparatus. 2. The mixed signal according to claim 1, wherein the calculation unit calculates the combination based on a spectrum of a second harmonic component in a mixed component of the plurality of first and second original signal data. Separation / extraction equipment. The sampling means samples the received signals for each of the plurality of sensors in a connected period obtained by connecting the first and second sampling periods, and acquires long-term sample data.
The separation means generates a plurality of third original signal data in the connected period by independent component analysis using the long-term sample data,
The calculation means calculates a correlation value between the plurality of first and second original signal data and the third original signal data, and calculates the combination based on the result. 2. The mixed signal separating / extracting device according to 1. The sampling means samples the received signals for each of the plurality of sensors in a connected period obtained by connecting the first and second sampling periods, and acquires long-term sample data.
The separation means generates a plurality of third original signal data in the connected period by independent component analysis using the long-term sample data,
2. The calculation unit compares the frequencies of the plurality of first and second original signal data and the third original signal data, and calculates the combination based on the result. The mixed signal separating / extracting device described in 1. The sampling means samples the received signals for each of the plurality of sensors in a connected period obtained by connecting the first and second sampling periods, and acquires long-term sample data.
The separation means generates a plurality of third original signal data in the connected period by independent component analysis using the long-term sample data,
The calculation means calculates a fourth-order statistic of the plurality of first and second original signal data and the third original signal data, and calculates the combination based on the result. 2. The mixed signal separating / extracting device according to 1. The sampling means samples the received signals for each of the plurality of sensors in a connected period obtained by connecting the first and second sampling periods, and acquires long-term sample data.
The separation means generates a plurality of third original signal data in the connected period by independent component analysis using the long-term sample data,
The calculation means calculates the combination based on a spectrum of a second harmonic component in a mixed component of the plurality of first and second original signal data and the third original signal data. 2. The mixed signal separating / extracting device according to 1. The calculation means generates all combinations of the plurality of first original signal data and the plurality of second original signal data, and the first and second original signal data included in each of the combinations are mutually 10. The mixed signal separating / extracting apparatus according to claim 2, wherein the combinations to be connected to each other are extracted from the correlation.

Images