Signal extracting apparatus
A signal extracting apparatus is provided based on independent component analysis with reference for a single measured signal as a signal processing technique that allows stable and quick extraction of a target signal from single measured signal even in a high-noise environment with a high noise ratio against a target signal to be extracted. The signal extracting apparatus includes: a single-signal measuring unit to measure a measured signal as a single time-series signal containing a target signal obtained by measurement by a single channel; a reference signal creating unit ( 2 ) to create a reference signal provided with timing-axis information based on the time-series signal obtained by the single-signal measuring unit; and a main processing unit ( 5 ) to perform independent component analysis with reference for a single measured signal by capturing the reference signal obtained by the reference signal creating unit ( 2 ) and providing the timing-axis information for an algorithm in a reference system to extract an independent component.
1. A signal extracting apparatus, comprising:
a signal input unit to input one or more time-series signals in which elements are arranged in temporal order;
a generating unit to generate a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
a reference signal creating unit to create a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
an independent component analysis with reference (ICA-R) unit to perform independent component analysis with reference for a plurality of the short time-series signals generated by the generating unit based on the reference signal to extract the independent signal,
wherein the independent signal to be extracted contains a signal that is considered to be autoregressive, and
the generating unit generates a plurality of the short time-series signals having different but partially overlapping temporal ranges so that one or more short time-series signals are dependent on one or more other short time-series signals when interpreted based on the autoregressive relationship.
2. The signal extracting apparatus according to claim 1 , wherein a single time-series signal is input by the signal input unit.
3. The signal extracting apparatus according to claim 1 , wherein the reference signal creating unit uses a temporal filter to generate, from the time-series signal, the reference signal indicating the temporal characteristics of the independent component to be extracted.
4. The signal extracting apparatus according to claim 1 , wherein the time-series signal is obtained by diagnosis of a living body, and the independent signal to be extracted includes a biosignal.
5. The signal extracting apparatus according to claim 4 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
6. A signal extracting apparatus, comprising:
(a) a single-signal input unit to input a single time-series signal;
(b) a reference signal creating unit to create a reference signal; and
(c) a main processing unit to perform independent component analysis with reference for a single measured signal by capturing the reference signal obtained by the reference signal creating unit and providing timing-axis information for an algorithm in a reference system to extract an independent component.
7. The signal extracting apparatus according to claim 6 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
8. A signal extracting program for a computer, comprising the steps of:
inputting one or more time-series signals in which elements are arranged in temporal order;
generating a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
creating a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
performing independent component analysis with reference for a plurality of the short time-series signals generated at the generating step based on the reference signal to extract the independent signal,
wherein the independent signal to be extracted contains a signal that is considered to be autoregressive, and
the generating unit generates a plurality of the short time-series signals having different but partially overlapping temporal ranges so that one or more short time-series signals are dependent on one or more other short time-series signals when interpreted based on the autoregressive relationship.
9. The signal extracting program according to claim 8 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
10. A signal extracting method for a computer, comprising the steps of:
inputting one or more time-series signals in which elements are arranged in temporal order;
generating a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
creating a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
performing independent component analysis with reference for a plurality of the short time-series signals generated at the generating step based on the reference signal to extract the independent signal,
wherein the independent signal to be extracted contains a signal that is considered to be autoregressive, and
the generating step generates a plurality of the short time-series signals having different but partially overlapping temporal ranges so that one or more short time-series signals are dependent on one or more other short time-series signals when interpreted based on the autoregressive relationship.
11. The signal extracting method according to claim 10 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
12. A signal extracting apparatus, comprising:
a signal input unit to input one or more time-series signals in which elements are arranged in temporal order;
a generating unit to generate a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
a reference signal creating unit to create a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
an independent component analysis with reference (ICA-R) unit to perform independent component analysis with reference for a plurality of the short time-series signals generated by the generating unit based on the reference signal to extract the independent signal,
wherein the generating step generates m short time-series signals having n elements to generate a matrix X of elements n×m, and
the ICA-R unit performs the independent component analysis with reference for the matrix X to calculate an independent component matrix Y of elements m×n having a plurality of independent signals as the elements and a matrix W of elements m×m in a matrix algebra relationship of Y=WX.
13. The signal extracting apparatus according to claim 12 , wherein the a single time-series signal is input by the signal input unit.
14. The signal extracting apparatus according to claim 12 , wherein the reference signal creating unit uses a temporal filter to generate, from the time-series signal, the reference signal indicating the temporal characteristics of the independent component to be extracted.
15. The signal extracting apparatus according to claim 12 , wherein the time-series signal is obtained by diagnosis of a living body, and the independent signal to be extracted includes a biosignal.
16. The signal extracting apparatus according to claim 15 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
17. A signal extracting program for a computer, comprising the steps of:
inputting one or more time-series signals in which elements are arranged in temporal order;
generating a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
creating a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
performing independent component analysis with reference (ICA-R) for a plurality of the short time-series signals generated at the generating step based on the reference signal to extract the independent signal,
wherein the generating step generates m short time-series signals having n elements to generate a matrix X of elements n×m, and
the ICA-R performing step performs the independent component analysis with reference for the matrix X to calculate an independent component matrix Y of elements m×n having a plurality of independent signals as the elements and a matrix W of elements m×m in a matrix algebra relationship of Y=WX.
18. The signal extracting program according to claim 17 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.
19. A signal extracting method for a computer, comprising the steps of:
inputting one or more time-series signals in which elements are arranged in temporal order;
generating a plurality of short time-series signals by selecting the elements from the time-series signals, wherein at least one time-series signal generates two or more short time-series signals where the selected elements have different but partially overlapping temporal ranges with each other;
creating a reference signal indicating a temporal characteristic of an independent signal to be extracted; and
performing independent component analysis with reference (ICA-R) for a plurality of the short time-series signals generated at the generating step based on the reference signal to extract the independent signal,
wherein the generating step generates m short time-series signals having n elements to generate a matrix X of elements n×m, and
the ICA-R performing step performs the independent component analysis with reference for the matrix X to calculate an independent component matrix Y of elements m×n having a plurality of independent signals as the elements and a matrix W of elements m×m in a matrix algebra relationship of Y=WX.
20. The signal extracting method according to claim 19 , wherein the independent signal to be extracted includes a signal regarding fetal heart sound.