CN100495388C - Signal processing method using space coordinates convert for realizing signal separation - Google Patents
Signal processing method using space coordinates convert for realizing signal separation Download PDFInfo
- Publication number
- CN100495388C CN100495388C CNB2006100630707A CN200610063070A CN100495388C CN 100495388 C CN100495388 C CN 100495388C CN B2006100630707 A CNB2006100630707 A CN B2006100630707A CN 200610063070 A CN200610063070 A CN 200610063070A CN 100495388 C CN100495388 C CN 100495388C
- Authority
- CN
- China
- Prior art keywords
- signal
- vector
- matrix
- processing method
- coordinate conversion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Abstract
The present invention discloses a signal processing method which realizes signal separation by utilizing space coordinate transformation. The method is used to process multi-channel observation signals which are X1(t)...xm (t) (m is more than 1), namely X(t). Firstly, an arbitrary group of orthogonal base vectors (E) of n-dimensional vector space, namely e1(t)...en(t), is sought, and the n-dimensional vector space is spanned by observation vectors X(t); then, according to the dimension of the vector space, coordinate transformation matrix M in the space is sought; cost function F(M) is built by using the determined transformation matrix M according to the independent characteristics of statistics; the base vectors E is transformed by utilizing the determined transformation matrix M, which determines to have the cost function get extremal transformation matrix Mm; a vector Em is got by utilizing the transformation matrix Mm to execute coordinate transformation to the base vectors; the vector Em and a source signal vector S(t) have the same starting point and are parallel to each other, which is the estimation of the standardized source signal vector S(t).
Description
Technical field
The present invention relates to the signal Processing field, more precisely to having mixed a plurality of statistics independently extraction of carrying out source signal of the hyperchannel observation signal of source signal or the disposal route of separation.
Background technology
In the system that a plurality of observation passages are arranged, add up independently often between the source signal of composition observation signal, and comprised the signal of expectation in the measurement in these source signals.In order from observation signal, to extract or isolate wanted signal, separate based on the statistics independent feature and can obtain good effect.But separation method commonly used at present is not very desirable to the structure theory of separation matrix (be also referred to as and separate mixed matrix), the value that the decomposition result of expecting for the extreme value realization of determining cost function when Data Update need obtain a lot of cost functions could realize, and the building mode of split-matrix is all different when decomposing at every turn, therefore the acquisition that the value of these cost functions can only be by each decomposition important statistic determine that this need pay sizable computing cost.When utilizing the Givens matrix that high dimensional signal is decomposed, need be to carrying out repeatedly Givens scanning in any two row in the split-matrix, but also need repeatedly repeat to obtain final separating resulting to identical step, this operand obviously is very huge.Though adopt adaptive method can reduce some operands, when but in case the element of hybrid matrix changes, just needing a very long time stablizes, if the element of hybrid matrix often changes (blending ratio of interference source in different channels in measuring such as arterial oxygen saturation may continue the very short time), adaptive decomposition method may all obtain the better estimation of split-matrix always.Therefore, at present used separation method is difficult to realize real-time and accurately to decomposition between the independent source signal or extraction, and particularly in some performances were not good especially embedded system, these methods almost can not be accomplished.
Summary of the invention
The purpose of this invention is to provide a kind of split-matrix method of passing through to make up, the decomposition with very low computing cost realization source signal particularly can realize the method for in real time signal being decomposed exactly in embedded system under the lower situation of signal dimension.
The technical scheme that realizes the foregoing invention purpose is as follows:
A kind of signal processing method that utilizes space coordinate conversion to realize Signal Separation, this method is used for multichannel observation signal x
1(t) ... x
m(t), m wherein〉1, be X (t) with the method representation of vector, handle; Observation signal vector X (t) is undertaken forming available following formulate after the linear superposition by source signal vector S (t):
X＝HS
Wherein, source signal vector S (t) is by one group of separate zero-mean signal s
1(t) ... s
n(t) form, wherein n 1, wherein at least one road signal is the signal of expectation;
H is the matrix of coefficients of linear superposition;
This method comprises following step:
At first, signal in orthogonal is decomposed, seek any one group of orthogonal basis vector e of the n-dimensional vector space of opening by observation signal vector X (t)
1(t) ... e
n(t), abbreviate base vector as, represent one group of base vector with E;
Then, according to the dimension of vector space, seek the transformation matrix of coordinates M in this space, the transform matrix M of the coordinate transform mode of determining in the vector space of determining has deterministic expression;
Utilize the transform matrix M of determining,, set up through cost function F (M) according to the statistics independent feature;
Utilize the transform matrix M of determining, base vector E is carried out coordinate transform, determine to make cost function to obtain the transform matrix M of extreme value
m
Utilize transform matrix M
mBase vector is carried out obtaining vectorial E after the coordinate transform
m, vectorial E
mAnd have identical starting point between the source signal vector S (t), and be to be parallel to each other, be the estimation of standardized source signal vector S (t).
Utilize determined vectorial E
m, determine each source signal component shared weight in observation signal.
Described observation signal is a physiological signal.
Described observation signal is the two paths of signals of gathering during arterial oxygen saturation is measured.
Described observation signal is the two paths of signals that has mixed mother's electrocardiosignal and fetus electrocardiosignal in the electrocardiogram acquisition simultaneously.
Utilize the matrix that obtains base vector and make cost function obtain the transformation matrix of coordinates of extreme value, obtain each the element s in the source signal vector S (t)
i(t) proportionate relationship in each observation signal.
Real-time acquisition two-way source signal s
1(t) and s
2(t) at two-way observation signal x
1(t) and x
2(t) the scale-up factor r in
1And r
2, and pass through r
1And r
2Probability in a period of time estimates to expect parameter.
It is to utilize signal matrix is carried out the realization of svd acquisition matrix that signal in orthogonal is decomposed.
It is to utilize the covariance matrix to signal to carry out the realization of quadrature decomposition acquisition matrix that signal in orthogonal is decomposed.
Described foundation through cost function be based on statistics independently the joint probability density function of each component to equal the marginal probability density function of each component long-pending.
Described foundation through cost function be based on statistics independently all the mutual cumulative amounts in k rank between the element of each component be zero, k＜∞ wherein, the exponent number of employed cumulative amount is 4 rank.
Described foundation through cost function be based on statistics independently the mutual information between each component be zero.
Utilize the expression formula of transform matrix M to realize statistic with base vector after the statistic denotation coordination conversion of the orthogonal basis vector before the coordinate transform.
The coordinate transform of described vector space, after transform matrix M is acted on base vector, after expression is rotated a certain angle to base vector, the coordinate under the reference system of former space.
Described transform matrix M is in a certain definite vector space, with one group of anglec of rotation θ
1..., θ
nMake up transform matrix M as parameter, one group of anglec of rotation represented by Θ, the anglec of rotation quantitaes of Θ in this space for n the required independent anglec of rotation of the completeness that realizes rotating.
The multivariate function F (Θ) that it is variable that described cost function F (M) changes into one group of anglec of rotation Θ.
The extreme value of described cost function F (Θ) for utilize the Θ enumerative technique, to find the solution F (Θ) be that the anglec of rotation Θ that zero system of equations that makes up, golden section search, method of steepest descent or Newton method obtain to make F (Θ) obtain extreme value determines to the first order derivative of Θ.
The port number m=2 of the observation signal of described observation signal vector X (t).
In the two-dimensional vector space, the quantity of anglec of rotation Θ is one, and cost function F (θ) is the function of a single variable of this anglec of rotation θ.
When observation signal upgrades, finding the solution F (θ) in real time is zero equation that makes up to the first order derivative of θ, acquisition makes F (θ) obtain the anglec of rotation θ of extreme value, and utilize this angle θ to make up rotation matrix, realize obtaining in real time with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
The present invention has following advantage:
1. the present invention's theory of having introduced space coordinate transformation in Signal Separation makes up split-matrix, has realized split-matrix expression formula determinacy in the identical vector space of dimension;
2. the present invention utilizes definite transformation matrix of coordinates to realize representing with the statistic of the base vector before the conversion statistic of the vector after the conversion, making only needs in the decomposable process to estimate a statistic, has avoided decomposing all and will reappraising the huge computing cost that is brought to the statistic of decomposing component in the middle of each;
3. the present invention utilizes the statistic of the base vector before the conversion to represent the statistic of the vector after the conversion, the function F (Θ) that it is variable that realization will change into one group of anglec of rotation Θ based on the cost function F (M) of the statistic of decomposing component, make each decomposition that definite cost function all be arranged, and can utilize the method acquisition of various solved function extreme values to make cost function function F (Θ) obtain the anglec of rotation Θ of extreme value, further reduced the computing cost;
4. the present invention has realized when the dimension of vector space hangs down, directly obtain to find the solution the expression formula that makes cost function function F (Θ) obtain the anglec of rotation Θ of extreme value, realize separating real-time and accurately or extracting each independent source, compare block data processing, greatly having enriched can be for the information of analyzing, with respect to adaptive algorithm, because do not need stabilization process from making decomposition result accurately instant more;
5. for the situation that need estimate to expect parameter by the ratio of reference source signal in different observation signals, the present invention is the split-matrix M of structure in real time
mThese scale-up factors can be provided in real time, greatly increase and analyzed used quantity of information, realize and to have carried out probability statistics to these scale-up factors, thereby make the acquisition of expectation parameter more reliable.
Description of drawings
Fig. 1 is the block diagram of the signal processing system of the embodiment of the invention 1;
Fig. 2 is the two-way observation signal that contains interference of actual measurement in the embodiment of the invention 17;
Fig. 3 is the separation signal among the Fig. 2 that obtains by method in the embodiment of the invention 17, the oscillogram of separation signal y1 (t) for disturbing wherein, and the oscillogram of separation signal y2 (t) to be the interested pulse that extracts under jamming pattern beat signal;
Fig. 4 is the pulse wave signal that obtains by method in the embodiment of the invention 20 and the scale-up factor of undesired signal in two-way observation signal trend over time, R1 represents to disturb the scale-up factor trend over time in the two-way observation signal, and R2 represents that the pulse scale-up factor (this coefficient can characterize arterial oxygen saturation) of signal in the two-way observation signal of beating scheme over time.
Embodiment
The present invention will be further described below in conjunction with the drawings and specific embodiments.
Embodiment 1
As shown in Figure 1, a kind of signal processing system of implementing the inventive method, observation signal receiver 1 can receive multichannel observation signal vector X (t).Observation signal vector X (t) is undertaken forming after the linear superposition by source signal vector S (t), and at least one road signal is the signal of expectation among the S (t), and the relation between observation signal and the source signal can be represented with following formula:
X＝HS
Wherein, source signal vector S (t) is by one group of separate zero-mean signal s
1(t) ... s
n(t) (n〉1) to form, wherein at least one road signal is the signal of expectation;
Observe vector by X (t) by multichannel observation signal x
1(t) ... x
m(t) (m 〉=n)
H is the matrix of coefficients of linear superposition.
Observation signal receiver 1 will be gathered picked up signal by signal conditioner 2, carry out every processing such as straight, amplifications, by analog-digital converter 3 above-mentioned signal all is converted to digital signal then, the signal that conversion obtains carries out pre-service by pretreater 4, including but not limited to as bandpass filter to eliminate those undesired signals outside expected frequency range.
Estimate the number of independent source signal according to actual conditions, it is the dimension of vector space, determine one group of independence anglec of rotation Θ quantity n in order to realize that complete rotation is required in this space, and make up the expression formula of the rotating coordinate transformation matrix of this vector space with this group anglec of rotation.Utilize the expression formula of transformation matrix to realize representing the statistic of postrotational vector with the statistic of the base vector before the rotation, thus the function F (Θ) that it is variable that realization will change into one group of anglec of rotation Θ based on the cost function F (M) of the statistic of postrotational vector.
The base vector that utilizes base vector estimator 5 to realize from observation signal vector vector space X (t) in the system utilizes extremum detector 7 to obtain to make cost function obtain one group of anglec of rotation Θ of extreme value then
m, utilize the anglec of rotation Θ that obtains
mMake up the transformation matrix of coordinates of expectation, and by the rotation that coordinate converter 8 is realized base vector, obtain with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
Embodiment 2
Present embodiment is on the basis of embodiment 1, and it is to utilize signal matrix is carried out the realization of svd acquisition matrix that employing is decomposed signal in orthogonal.
Embodiment 3
Present embodiment is on the basis of embodiment 1, and it is to utilize the covariance matrix to signal to carry out the realization of quadrature decomposition acquisition matrix that employing is decomposed signal in orthogonal.
Present embodiment is on the basis of embodiment 1, and cost function is set up based on 4 rank cumulative amounts of postrotational vector.
Present embodiment is on the basis of embodiment 1, and employing is enumerated Θ and obtained to make F (Θ) obtain the anglec of rotation Θ of extreme value.
Embodiment 6
Present embodiment is on the basis of embodiment 1, and adopting and finding the solution F (Θ) is zero system of equations that makes up to the first order derivative of Θ, obtains to make F (Θ) obtain the anglec of rotation Θ of extreme value.
Embodiment 7
Present embodiment is on the basis of embodiment 1, adopts golden section search to obtain to make F (Θ) obtain the anglec of rotation Θ of extreme value.
Embodiment 8
Present embodiment is on the basis of embodiment 1, adopts method of steepest descent to obtain to make F (Θ) obtain the anglec of rotation Θ of extreme value.
Embodiment 9
Present embodiment is on the basis of embodiment 1, adopts Newton method to obtain to make F (Θ) obtain the anglec of rotation Θ of extreme value.
The two paths of signals x that a kind of signal processing method, observation signal are
1(t), x
2(t).Two paths of signals is that two paths of signals is by two independent source signal s
1(t) and s
2(t) linear being formed by stacking.Remove the precognition noise when carrying out pre-service, and after going average and removing phase differential, can represent with following formula between two-way observation signal and two independent source signals:
The method of utilizing any quadrature to decompose obtains split-matrix and realizes observation signal X (t) is decomposed into two orthogonal basis vector E (t) rotating coordinate transformation Givens rotation matrix M in the two-dimensional vector space
gRealize Givens rotation matrix M
gThe middle complete rotation that has realized 0～2 π with an anglec of rotation θ as parameter.With Givens rotation matrix M
gAct on base vector E (t) back and obtain (t) coordinate under former spatial frame of reference of postrotational vectorial E ', be expressed as:
E′(t)＝M
gE(t)
Because when θ rotates base vector in～2 π, rotation matrix M
gDeterministic expression is arranged, and just the value of expression formula changes along with the variation of anglec of rotation θ, so can utilize rotation matrix M
gExpression formula realize representing postrotational vectorial E ' statistic (E ' represents with ∑) (t) that then this relation can be expressed as with the statistic (E represents with ∑) of the vector E of the orthogonal basis before the rotation (t):
∑E′＝g(M
g，∑E)
In case and after base vector E (t) determined, all statistic ∑ E of base vector also determined thereupon, and the element of rotation matrix with anglec of rotation θ as parameter; Therefore also just can to regard as with anglec of rotation θ be the function of a single variable of independent variable to postrotational vectorial E ' statistic ∑ E ' (t), is expressed as ∑ E '=g (θ).
For the decomposition result that obtains to expect, utilize postrotational vectorial E ' cumulative amount (t) to set up cost function F (∑ E '), when cost function F (∑ E ') when obtaining maximum value, corresponding postrotational vectorial Em obtain with source signal vector S (t) between have identical starting point, and be parallel to each other.
As described above, because postrotational vectorial E ' statistic ∑ E ' (t) is to be the function of a single variable g (θ) of independent variable with anglec of rotation θ, be the function of a single variable F (θ) of independent variable so be expressed as that F (∑ E ') can be converted to equally with anglec of rotation θ.
Determine to make cost function to obtain the anglec of rotation θ of extreme value
m, utilize this anglec of rotation θ
mMake up rotation matrix M
g, and with this rotation matrix M
gAct on base vector, obtain with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
Embodiment 11
Present embodiment is on the basis of embodiment 10, and it is to utilize signal matrix is carried out the realization of svd acquisition matrix that employing is decomposed signal in orthogonal.
Embodiment 12
Present embodiment is on the basis of embodiment 10, and it is to utilize the covariance matrix to signal to carry out the realization of quadrature decomposition acquisition matrix that employing is decomposed signal in orthogonal.
Embodiment 13
Present embodiment is on the basis of embodiment 10, and cost function is set up based on 4 rank cumulative amounts of postrotational vector.
Embodiment 14
Present embodiment is on the basis of embodiment 10, and adopting and finding the solution F (θ) is zero equation that makes up to the first order derivative of θ, obtains to make F (θ) obtain the anglec of rotation θ of extreme value
mCalculation expression, quick when being implemented in Data Update, construct the rotation matrix M of expectation in real time
g, obtain simultaneously with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
Present embodiment is on the basis of embodiment 10,14, and this is quick, and the method for separation source signal is applied in the embedded system in real time.
Embodiment 16
Present embodiment is on the basis of embodiment 10, and the observation signal of processing is the two paths of signals that has mixed mother's electrocardiosignal and fetus electrocardiosignal in the electrocardiogram acquisition simultaneously.
Embodiment 17
Present embodiment is on the basis of embodiment 10,14, the observation signal of handling be the two paths of signals gathered during arterial oxygen saturation is measured (wherein, source signal is beat signal and a undesired signal of pulse), and realize separating in real time pulse beat signal and undesired signal with the method.
Fig. 3 has showed the decomposition result of this method in the process of decomposing in real time, wherein Fig. 2 has showed the oscillogram of the two-way observation signal that contains motion artifacts of an actual measurement, as can be seen under the strong jamming background, the beat waveform character of signal of interested pulse is covered fully.The oscillogram of separation signal y1 (t) for disturbing among Fig. 3, and the oscillogram of y2 (t) to be the interested pulse that extracts under jamming pattern beat signal, the signal that extracts as can be seen and the pulse signal of beating is consistent.
Embodiment 18
Present embodiment is on the basis of embodiment 10, utilizes the matrix that obtains base vector and makes cost function obtain the transformation matrix of coordinates of extreme value, determines each the element s in the source signal vector S (t)
i(t) proportionate relationship in each observation signal.
Embodiment 19
Present embodiment is on the basis of embodiment 14,18, real-time acquisition two-way source signal s
1(t) and s
2(t) at two-way observation signal x
1(t) and x
2(t) the scale-up factor r in
1And r
2, and pass through r
1And r
2Probability in a period of time is estimated some expectation parameter;
Embodiment 20
Present embodiment is on the basis of embodiment 17,19, utilize the probability of the scale-up factor of pulse wave signal in the two-way observation signal to estimate arterial oxygen saturation, represent to disturb scale-up factor trend over time in the two-way observation signal as R1 among Fig. 4, R2 represents the pulse scale-up factor of signal in the two-way observation signal of beating, this coefficient can characterize arterial oxygen saturation, trend over time.In 0～15s, can obtain scale-up factor continuously in real time as can be seen, if use block data processing, several scale-up factors that in the identical time period, generally can only obtain to disperse.Shown among the figure that scale-up factor (weight of each source signal is got in touch during directly with linear stack) is a real-time change, and adaptive algorithm needs a period of time stable when scale-up factor changes, and therefore adaptive algorithm is difficult to guarantee to obtain real-time and accurately the scale-up factor expected under the changing situation of this scale-up factor.Therefore compare the method for conventional block data processing, the real-time separation method that the present invention realizes has enriched greatly can be for the information of analysis and judgement, and compared to adaptive algorithm, the real-time separation method that the present invention realizes does not need stabilization time, can provide than adaptive algorithm more near the scale-up factor of expecting.
Embodiment 21
A kind of signal processing method, observation signal are three road signal x
1(t), x
2(t), x
3(t), three road signals are by three paths of independent source signal s
1(t), s
2(t) and s
3(t) linear being formed by stacking.Remove the precognition noise when carrying out pre-service, and after going average and removing phase differential, can represent with following formula between three road observation signals and three independent source signals:
The method of utilizing any quadrature to decompose obtains split-matrix and realizes observation signal X (t) is decomposed into three orthogonal basis vector E (t), and rotating coordinate transformation can adopt Eulerian angle Θ (by angle of nutation θ, angle of precession ψ, angle of rotation in the trivector space
Form) for parameter makes up rotation matrix, its expression formula is as follows:
Angle of nutation θ range of movement among the rotation matrix M be [0, π), angle of precession motion ψ scope be [0,2 π), angle of rotation
Range of movement be [0,2 π).Rotation matrix M is acted on base vector E (t) back obtains (t) coordinate under former spatial frame of reference of postrotational vectorial E ', be expressed as:
E′(t)＝ME(t)
Because when utilizing Eulerian angle Θ that base vector is rotated, rotation matrix M has deterministic expression, just the value of expression formula changes along with the variation of Eulerian angle Θ, so can utilize the expression formula of rotation matrix M to realize representing postrotational vectorial E ' statistic (E ' represents with ∑) (t) with the statistic (E represents with ∑) of the vector E of the orthogonal basis before the rotation (t), then this relation can be expressed as:
∑E′＝g(M，∑E)
In case and after base vector E (t) determined, all statistic ∑ E of base vector also determined thereupon, and the element of rotation matrix with Eulerian angle Θ as parameter; Therefore also just can to regard as with Eulerian angle Θ be the multivariate function of independent variable to postrotational vectorial E ' statistic ∑ E ' (t), is expressed as ∑ E '=g (Θ).
For the decomposition result that obtains to expect, utilize the mutual information minimum of postrotational vectorial E ' between (t) to set up cost function F (∑ E '), when cost function F (∑ E ') obtains minimizing the time corresponding postrotational vectorial E
mObtain with source signal vector S (t) between have identical starting point, and be parallel to each other.
As described above, because postrotational vectorial E ' statistic ∑ E ' (t) is to be the multivariate function g (Θ) of independent variable with Eulerian angle Θ, be the multivariate function F (Θ) of independent variable so be expressed as that F (∑ E ') can be converted to equally with Eulerian angle Θ.
Determine to make cost function to obtain minimizing Eulerian angle Θ
m, utilize these Eulerian angle Θ
mMake up rotation matrix M, and this rotation matrix M acted on base vector, obtain with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
Embodiment 22
Present embodiment is on the basis of embodiment 21, and it is to utilize signal matrix is carried out the realization of svd acquisition matrix that employing is decomposed signal in orthogonal.
Embodiment 23
Present embodiment is on the basis of embodiment 21, and it is to utilize the covariance matrix to signal to carry out the realization of quadrature decomposition acquisition matrix that employing is decomposed signal in orthogonal.
Embodiment 24
Present embodiment is on the basis of embodiment 21, adopts golden section search, Newton method, and method of steepest descent etc. are found the solution and made the Eulerian angle Θ of cost function F (Θ) when obtaining minimal value
m
Embodiment 25
Present embodiment is on the basis of embodiment 21, utilizes the method for space rotation to handle three tunnel mixed signals of (the disturbing as power frequency) of containing mother's electrocardiosignal, fetus electrocardiosignal and undesired signal, realizes the separation to three independent signals.
Claims (20)
1. signal processing method that utilizes space coordinate conversion to realize Signal Separation, this method is used for multichannel observation signal x
1(t) ... x
m(t), m wherein〉1, be X (t) with the method representation of vector, handle; Observation signal vector X (t) is undertaken forming available following formulate after the linear superposition by source signal vector S (t):
X＝HS
Wherein, source signal vector S (t) is by one group of separate zero-mean signal s
1(t) ... s
n(t) form, wherein n 1, wherein at least one road signal is the signal of expectation;
H is the matrix of coefficients of linear superposition;
This method comprises following step:
At first, signal in orthogonal is decomposed, seek any one group of orthogonal basis vector e of the n-dimensional vector space of opening by observation signal vector X (t)
1(t) ... e
n(t), abbreviate base vector as, represent one group of base vector with E;
Then, according to the dimension of vector space, seek the transformation matrix of coordinates M in this space, the transform matrix M of the coordinate transform mode of determining in the vector space of determining has deterministic expression;
Utilize the transform matrix M of determining,, set up through cost function F (M) according to the statistics independent feature;
Utilize the transform matrix M of determining, base vector E is carried out coordinate transform, determine to make cost function to obtain the transform matrix M of extreme value
m
Utilize transform matrix M
mBase vector is carried out obtaining vectorial E after the coordinate transform
m, vectorial E
mAnd have identical starting point between the source signal vector S (t), and be to be parallel to each other, be the estimation of standardized source signal vector S (t).
2. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1 is characterized by and utilizes determined vectorial E
m, determine each source signal component shared weight in observation signal.
3. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1, it is characterized by described observation signal is physiological signal.
4. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 3, it is characterized by described observation signal is the two paths of signals of gathering during arterial oxygen saturation is measured.
5. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 3, it is characterized by described observation signal is the two paths of signals that has mixed mother's electrocardiosignal and fetus electrocardiosignal in the electrocardiogram acquisition simultaneously.
6. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1 is characterized by and utilizes the matrix that obtains base vector and make cost function obtain the transformation matrix of coordinates of extreme value, obtains each the element s in the source signal vector S (t)
i(t) proportionate relationship in each observation signal.
7. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 6 is characterized by real-time acquisition two-way source signal s
1(t) and s
2(t) at two-way observation signal x
1(t) and x
2(t) the scale-up factor r in
1And r
2, and pass through r
1And r
2Probability in a period of time estimates to expect parameter.
8. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1, it is characterized in that signal in orthogonal is decomposed is to utilize signal matrix is carried out the realization of svd acquisition matrix.
9. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1 is characterized by that signal in orthogonal is decomposed is to utilize covariance matrix to signal to carry out quadrature to decompose and obtain matrix and realize.
10. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1, it is characterized by described foundation through cost function be based on statistics independently the joint probability density function of each component to equal the marginal probability density function of each component long-pending.
11. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 10, it is characterized by described foundation through cost function be based on statistics independently all the mutual cumulative amounts in k rank between the element of each component be zero, k＜∞ wherein, the exponent number of employed cumulative amount is 4 rank.
12. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 10, it is characterized by described foundation through cost function be based on statistics independently the mutual information between each component be zero.
13. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1 is characterized by the expression formula of utilizing transform matrix M and realizes statistic with base vector after the statistic denotation coordination conversion of the orthogonal basis vector before the coordinate transform.
14. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1, it is characterized by the coordinate transform of described vector space, after transform matrix M is acted on base vector, after expression is rotated a certain angle to base vector, the coordinate under the reference system of former space.
15., it is characterized by described transform matrix M in a certain definite vector space, with one group of anglec of rotation θ according to claim 13 or the 14 described signal processing methods that utilize space coordinate conversion to realize Signal Separation
1..., θ
nMake up transform matrix M as parameter, one group of anglec of rotation represented by Θ, the anglec of rotation quantitaes of Θ in this space for n the required independent anglec of rotation of the completeness that realizes rotating.
16. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 15 is characterized by described cost function F (M) multivariate function F (Θ) that to change into one group of anglec of rotation Θ be variable.
17. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 16, the extreme value that it is characterized by described cost function F (Θ) for utilize the Θ enumerative technique, to find the solution F (Θ) be that the anglec of rotation Θ that zero system of equations that makes up, golden section search, method of steepest descent or Newton method obtain to make F (Θ) obtain extreme value determines to the first order derivative of Θ.
18. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 1 is characterized by the port number m=2 of the observation signal of described observation signal vector X (t).
19. according to claim 1 or the 18 described signal processing methods that utilize space coordinate conversion to realize Signal Separation, it is characterized by in the two-dimensional vector space, the quantity of anglec of rotation Θ is one, cost function F (θ) is the function of a single variable of this anglec of rotation θ.
20. the signal processing method that utilizes space coordinate conversion to realize Signal Separation according to claim 19, it is characterized in that, when observation signal upgrades, finding the solution F (θ) in real time is zero equation that makes up to the first order derivative of θ, acquisition makes F (θ) obtain the anglec of rotation θ of extreme value, and utilize this angle θ to make up rotation matrix, realize obtaining in real time with source signal vector S (t) between have identical starting point, and the vectorial E that is parallel to each other
m
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100630707A CN100495388C (en) | 2006-10-10 | 2006-10-10 | Signal processing method using space coordinates convert for realizing signal separation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100630707A CN100495388C (en) | 2006-10-10 | 2006-10-10 | Signal processing method using space coordinates convert for realizing signal separation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101162453A CN101162453A (en) | 2008-04-16 |
| CN100495388C true CN100495388C (en) | 2009-06-03 |
Family
ID=39297377
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2006100630707A Active CN100495388C (en) | 2006-10-10 | 2006-10-10 | Signal processing method using space coordinates convert for realizing signal separation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100495388C (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101814194B (en) * | 2010-04-08 | 2012-08-22 | 合肥工业大学 | Rotation angle increment method for eliminating singularity of Eulerian angle |
| US9959247B1 (en) | 2017-02-17 | 2018-05-01 | Google Llc | Permuting in a matrix-vector processor |
| CN109413543B (en) * | 2017-08-15 | 2021-01-19 | 音科有限公司 | Source signal extraction method, system and storage medium |
| CN109494712A (en) * | 2018-10-31 | 2019-03-19 | 国网辽宁省电力有限公司 | A kind of low-frequency oscillation of electric power system parameter identification method |
| CN112472079B (en) * | 2020-11-23 | 2022-06-21 | 青岛歌尔智能传感器有限公司 | Blood oxygen saturation detection device, equipment and storage medium |
-
2006
- 2006-10-10 CN CNB2006100630707A patent/CN100495388C/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN101162453A (en) | 2008-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100495388C (en) | Signal processing method using space coordinates convert for realizing signal separation | |
| Rouquette et al. | Estimation of frequencies and damping factors by two-dimensional ESPRIT type methods | |
| Särkkä | Recursive Bayesian inference on stochastic differential equations | |
| CN105589056B (en) | A kind of multiple target distance field mixing source localization method | |
| JP4271041B2 (en) | Blind source separation using spatial fourth-order cumulant matrix bundle | |
| CN100460893C (en) | Method for analyzing noise elimination of earthquake based on independent components in Pearson system | |
| CN106125041A (en) | The wideband source localization method of sparse recovery is weighted based on subspace | |
| Filzmoser | Robust principal component and factor analysis in the geostatistical treatment of environmental data | |
| US7706478B2 (en) | Method and apparatus of source separation | |
| JP5323860B2 (en) | Method for separating a mixed signal into a plurality of component signals | |
| CN109975842B (en) | Wavelet transform-based Beidou satellite signal high-precision blind capturing method | |
| Roemer | Advanced algebraic concepts for efficient multi-channel signal processing | |
| Stadnitski | Some critical aspects of fractality research | |
| Ramsey | Regression over timescale decompositions: a sampling analysis of distributional properties | |
| Liu et al. | A general-purpose signal processing algorithm for biological profiles using only first-order derivative information | |
| Guerrier et al. | Wavelet-based improvements for inertial sensor error modeling | |
| Kleiber et al. | Modern methods of signal processing applied to gyrokinetic simulations | |
| US20120162004A1 (en) | Apparatus and method for estimating direction of arrival of signal in communication system | |
| Brasseur | Ocean data assimilation using sequential methods based on the Kalman filter: From theory to practical implementations | |
| Hoffman | The effect of thinning and superobservations in a simple one-dimensional data analysis with mischaracterized error | |
| CN110672127B (en) | Real-time calibration method for array type MEMS magnetic sensor | |
| Washizawa et al. | A flexible method for envelope estimation in empirical mode decomposition | |
| CN115685067A (en) | Normal-mode signal blind estimation method and system for positioning and tracking of multi-rotor unmanned aerial vehicle | |
| CN107179521B (en) | A kind of Robust Subspace estimation method of low complex degree | |
| CN112834847A (en) | Radiation EMI noise standard exceeding analysis method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C56 | Change in the name or address of the patentee |
Owner name: EDAN INSTRUMENTS, INC. Free format text: FORMER NAME: EDAN INSTRUMENTS, CO., LTD |
|
| CP01 | Change in the name or title of a patent holder |
Address after: 518054, fifth, new energy building, 2009 Nanhai Road, Shenzhen, Guangdong, Nanshan District (elevator No. four) Patentee after: Shenzhen Edan Precision Instruments, Inc. Address before: 518054, fifth, new energy building, 2009 Nanhai Road, Shenzhen, Guangdong, Nanshan District (elevator No. four) Patentee before: Shenzhen Libang Precision Instrument Co., Ltd. |
|
| C56 | Change in the name or address of the patentee | ||
| CP02 | Change in the address of a patent holder |
Address after: 518067 Guangdong city of Shenzhen province Nanshan District Nanhai Road No. 1019 Nanshan Park building B on the third floor of medical devices Patentee after: Shenzhen Edan Precision Instruments, Inc. Address before: 518054, fifth, new energy building, 2009 Nanhai Road, Shenzhen, Guangdong, Nanshan District (elevator No. four) Patentee before: Shenzhen Edan Precision Instruments, Inc. |