CN104182645A - Empirical mode decomposition and sliding time window weighted least square method based brain-computer interface extraction method - Google Patents

Abstract

The invention relates to a signal extraction method, particularly relates to an empirical mode decomposition and sliding time window weighted least square method based brain-computer interface extraction method. The empirical mode decomposition and sliding time window weighted least square method based brain-computer interface extraction method aims at solving the problems that the existing physiological interference seriously influences accurate extraction of brain-computer signals and limits development of the brain-computer interface technology. The empirical mode decomposition and sliding time window weighted least square method based brain-computer interface extraction method comprises step one, detecting at a brain tissue position; step two, detecting to obtain electrical signals; step three, measuring delta [HbO2]<N>(k) and delta[HHb]<N>(k) through a detector D1 and measuring delta[HbO2]<F>(k) and delta[HHb]<F>(k) through a detector D2; step four, expressing the delta [HbO2]<F>(k) or the delta[HHb]<N>(k) through x(k); step five, expressing the delta[HbO2]<F>(k) or the delta[HHb]<F>(k) through y (k); step six, enabling the y(k) to contain r(k) and i(k) and expressing the physiological interference as the following formula; step seven, calculating an expression of brain functional signals according to a formula that y(k) is equal to r(k) plus i(k) and the flowing formula; step eight, obtaining the brain functional signals s(k). The empirical mode decomposition and sliding time window weighted least square method based brain-computer interface extraction method is applied to the signal processing field.

Description

Based on the brain-computer interface method for extracting signal of empirical mode decomposition and time slip-window weighted least-squares method

Technical field

The present invention relates to a kind of method for extracting signal, be specifically related to the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method.

Background technology

Near-infrared spectrum technique can be used for oxyhemoglobin concentration change Δ [HbO in cerebral cortex ₂] and the signal measurement of reduced hemoglobin concentration change Δ [HHb], further expand brain-computer interface research.But, the application of near-infrared spectrum technique is limited by the impact of the physiological activity of human body always, is referred to as physiology and disturbs, particularly in the time that brain tissue shows very strong nonuniformity, this physiology serious interference affects the accurate extraction of brain machine signal, the development of restriction brain-computer interface technology.

Summary of the invention

The present invention seeks to affect for solving current physiology serious interference the accurate extraction of brain machine signal, the development of restriction brain-computer interface technology.And the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method is proposed.

Above-mentioned goal of the invention is achieved through the following technical solutions:

Step 1, the near-infrared probe that uses light source S and two detecting device D1 and D2 to form in brain tissue to be measured position are surveyed;

Step 2, detection diffuse reflection light intensity signal also convert the electric signal that obtains reaction intensity signal through photoelectric sensor;

Step 3, electric signal are obtained the time series △ [HbO of the oxyhemoglobin concentration change amount that detecting device D1 records by revising langbobier law ₂] ⁿand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ⁿ(k) the time series △ [HbO of the oxyhemoglobin concentration change amount that, detecting device D2 records ₂] ^fand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ^f(k);

Step 4, use x (k) represent △ [HbO ₂] ⁿor △ [HHb] (k) ⁿ(k);

Step 5, use y (k) represent △ [HbO ₂] ^for △ [HHb] (k) ^f(k);

In step 6, y (k), comprise cerebration signal r (k) and physiology and disturb i (k), i.e. y (k)=r (k)+i (k), physiology disturbs and can be expressed as

$\hat{i}\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=k-M}{\overset{k+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right);$

Step 7, according to y (k)=r (k)+i (k) and can extrapolate the expression formula of brain function signal;

Step 8, utilize weighted least square algorithm as cost function, ask for optimized coefficients w _i,j(k), then by the optimized coefficients w asking for _i,j(k) bring the expression formula of brain function signal into, can obtain brain function signal s (k).

Invention effect

The inventive method is on the basis of multiple spurs measuring method, the physiology interference that the hemodynamic parameter that near-end detecting device D1 is obtained and remote sensor D2 are subject to has correlativity and takes into full account, estimate the size of physiology undesired signal by building two-dimensional optimization model, use moving window to be optimized for physiology interfere with dynamic characteristic, wherein two dimension refers to respectively near-end detection signal to utilize the IMF component of empirical mode decomposition algorithm acquisition and the time series of each IMF component simultaneously.In the rejecting of fully estimating to carry out on the basis that physiology disturbs physiology interference, thereby realize the object that brain-computer interface signal is accurately extracted.

Brief description of the drawings

Fig. 1 is process flow diagram of the present invention;

Fig. 2 is the Near-infrared Brain machine interface signal detection probe structure based on multiple spurs measuring method, and wherein a represents scalp, and b represents skull, and c represents cerebrospinal fluid, and d represents ectocinerea, and e represents white matter of brain.

Embodiment

Embodiment one: the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method of present embodiment, specifically prepare according to following steps:

Step 1, the near-infrared probe that uses light source S and two detecting device D1 and D2 to form in brain tissue to be measured position are surveyed;

Step 2, detection diffuse reflection light intensity signal also convert the electric signal that obtains reaction intensity signal through photoelectric sensor;

Step 3, electric signal are obtained the time series △ [HbO of the oxyhemoglobin concentration change amount that detecting device D1 records by revising langbobier law ₂] ⁿand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ⁿ(k) the time series △ [HbO of the oxyhemoglobin concentration change amount that, detecting device D2 records ₂] ^fand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ^f(k); Described langbobier law is for transmission, revise langbobier law for be diffuse reflection;

Step 4, use x (k) represent △ [HbO ₂] ⁿor △ [HHb] (k) ⁿ(k);

Step 5, use y (k) represent △ [HbO ₂] ^for △ [HHb] (k) ^f(k);

In step 6, y (k), comprise cerebration signal r (k) and physiology and disturb i (k), i.e. y (k)=r (k)+i (k), physiology disturbs and can be expressed as

$\hat{i}\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=k-M}{\overset{k+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right);$

Step 7, according to y (k)=r (k)+i (k) and can extrapolate the expression formula of brain function signal;

Step 8, utilize weighted least square algorithm as cost function, ask for optimized coefficients w _i,j(k), then by the optimized coefficients w asking for _i,j(k) bring the expression formula of brain function signal into, can obtain brain function signal s (k).

Embodiment two, present embodiment are different from embodiment one: the light source S described in step 1 adopts integrated dual wavelength near-infrared light source, and light source S is r to the air line distance between near-end detecting device D1 ₁; Light source S is r to the air line distance between remote sensor D2 ₂.

Other step and parameter are identical with embodiment one.

Embodiment three: present embodiment is different from embodiment one or two: two kinds of wavelength that described dual wavelength near-infrared light source sends are respectively λ ₁=750nm, λ ₂=830nm.

Other step and parameter are identical with embodiment one or two.

Embodiment four: present embodiment is different from one of embodiment one to three: the spacing r of described light source S and detecting device D1 ₁for 10mm, the spacing r of light emitting source S and detecting device D2 ₂for 35mm.

Other step and parameter are identical with one of embodiment one to three.

Embodiment five: present embodiment is different from one of embodiment one to four: △ [HbO in described step 3 ₂] ⁿand △ [HHb] (k) ⁿ(k) be:

$\mathrm{\&Delta;}{\left[{\mathrm{HbO}}_2\right]}^N\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^N\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^N\left(k\right)/\mathrm{DPF})}{r_1({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

$\mathrm{\&Delta;}{\left[\mathrm{HHb}\right]}^N\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^N\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{H{\mathrm{bO}}_2}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^N\left(k\right)/\mathrm{DPF})}{r_1({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

△ [HbO ₂] ^fand △ [HHb] (k) ^f(k) be:

$\mathrm{\&Delta;}{\left[{\mathrm{HbO}}_2\right]}^F\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^F\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^F\left(k\right)/\mathrm{DPF})}{r_2({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

$\mathrm{\&Delta;}{\left[\mathrm{HHb}\right]}^F\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^F\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{H{\mathrm{bO}}_2}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^F\left(k\right)/\mathrm{DPF})}{r_2({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

Wherein, ε _hHb(λ ₁) for optical source wavelength be λ ₁time the extinction coefficient of reduced hemoglobin, ε _hHb(λ ₂) for optical source wavelength be λ ₂time the extinction coefficient of reduced hemoglobin,

(λ ₁) for optical source wavelength be λ ₁time oxyhemoglobin extinction coefficient, (λ ₂) for optical source wavelength be λ ₂time oxyhemoglobin extinction coefficient,

with be illustrated in S-D1 measurement passage and wavelength and be λ respectively ₁and λ ₂time absorbance variable quantity time series, with be illustrated in S-D2 measurement passage and wavelength and be respectively λ ₁and λ ₂time absorbance variable quantity time series,

K is the time, k=1, and 2 ..., N; N is positive integer;

DPF is the differential path factor.

Other step and parameter are identical with one of embodiment one to four.

Embodiment six: present embodiment is different from one of embodiment one to five: x in described step 4 (k) utilizes empirical mode decomposition algorithm that x (k) is decomposed into N solid-state mode function component IMF component, using residual components as last IMF component, x (k) is expressed as

$x\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{c}_i\left(k\right)$

Wherein, c _i(k) the IMF component for decomposing; N is positive integer.

Other step and parameter are identical with one of embodiment one to five.

Embodiment seven: present embodiment is different from one of embodiment one to six: y in described step 6 (k)=r (k)+i (k), for specific time point k, select the time window [k-M of the 2M+1 length centered by this time point, k-M+1, k+M], utilize two-dimentional weight coefficient w _i,j(k) relation that IMF component and physiology disturb is described;

Described in step 6 wherein for the estimation of i (k), i=1,2 ..., N, w _i,j(k) be optimized coefficients, c _i(j) i empirical modal component of expression is in the size in j moment, and wherein, M is half window length, and k is positive integer, and N is positive integer.

Other step and parameter are identical with one of embodiment one to six.

Embodiment eight: present embodiment is different from one of embodiment one to seven: the expression formula of brain function signal in described step 7:

$S\left(k\right)=y\left(k\right)-\hat{i}\left(k\right)=r\left(k\right)+[i\left(k\right)-\hat{i}\left(k\right)]$

Wherein, s (k) is brain function signal, and r (k) is the estimation of brain function signal s (k), and i (k) is physiology interference, for the estimation of i (k).

Other step and parameter are identical with one of embodiment one to seven.

Embodiment nine: present embodiment is different from one of embodiment one to eight: utilize in described step 8, obtain brain function signal s (k):

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{s}^2\left(n\right)$

Further be expressed as

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]}^2$

Wherein, J (k) is square error performance function; χ is exponential weighting factor, χ=0.99, and n=1 ... k, k is positive integer, i=1,2 ..., N, N is positive integer, M is half window length, wi, _j(k) be i IMF component at the weight coefficient optimized coefficients in j moment, c _i(j) i empirical modal component of expression is at the size in j moment, s ²(n) be moment n brain function signal s (n) square, y (n) obtains brain function signal for moment n utilizes S-D2 passage, solves the w that makes J (k) minimum _i,j(k), obtain brain function signal s (k).

Other step and parameter are identical with one of embodiment one to eight.

Embodiment ten: present embodiment is different from one of embodiment one to nine: in described step 8, the preparation method of brain function signal s (k) is:

One, accumulative total square error performance function J (k) minimum that represents to make error brain function signal s (k) by weighted least square algorithm estimation criterion, J (k) is expressed as:

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{s}^2\left(n\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]}^2$

W in formula _i,j(k) be optimized coefficients, c _i(j) i empirical modal component of expression is in the size in j moment, and y (n) obtains brain function signal for moment n utilizes S-D2 passage, n=1 ... k, k is positive integer, i=1,2 ..., N, N is positive integer, M is half window length;

Two, solve optimization coefficient w _i,j(k):

By to J (k) with respect to w _i,j(k) differentiate, and make it equal zero,

$-2\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]c_l\left(m\right)=0$

W in formula _i,j(k) be i empirical modal component at the weight coefficient in j moment, c _i(j) i empirical modal component of expression is at the size in j moment, c _l(m) for representing the size of l empirical modal component in the m moment

Obtained by above formula

$\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}y\left(n\right)c_l\left(m\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{{\mathrm{\&chi;}}^{k-n}c}_i(n-j)c_l(n-m)$

Or

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=k-M}{\overset{k+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)R_{i,l;j,m}\left(k\right)=p_{l,m}\left(k\right),l=1,...,N,m=k-M,...,k+M$

In formula, k is positive integer, wherein, and p _l,mand R (k) _{i, l; j,m}(k) expression formula is

$p_{l,m}\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}y\left(n\right)c_l\left(m\right)$

$R_{i,l;j,m}\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{c}_i\left(j\right)c_l\left(m\right)$

In formula, k is positive integer, c _i(j) i empirical modal component of expression is at the size in j moment, c _l(m), for l empirical modal component of expression is in the size in m moment, y (n) obtains brain function signal for moment n utilizes S-D2 passage;

Being expressed as of its matrix form

Can further be reduced to

R(k)w(k)＝p(k)

If matrix R (k) is nonsingular, optimal coefficient calculates by following formula

w ^*(k)＝R ^-1(k)p(k)

Wherein, w ^*(k) be expressed as the optimum solution of w (k),

R ^-1(K) be the inverse matrix of R (K),

$w\left(k\right)=\left[\begin{array}{c}{w}_{1,-M}\left(k\right)\\ w_{1,-M+1}\left(k\right)\\ .\\ .\\ .\\ w_{N,M}\left(k\right)\end{array}\right],$

$p\left(k\right)=\left[\begin{array}{c}{p}_{1,-M}\left(k\right)\\ p_{1,-M+1}\left(k\right)\\ .\\ .\\ .\\ p_{N,M}\left(k\right)\end{array}\right],$

$c\left(k\right)=\left[\begin{array}{c}{c}_1(-M)\\ c_1(-M+1)\\ .\\ .\\ .\\ c_N\left(M\right)\end{array}\right];$

Three, solve brain function signal s (k):

s(k)＝y(k)-c ^T(k)w ^*(k)，

Wherein c ^t(k) what represent is the transposed matrix of c (k), w ^*(k) optimal coefficient vector that expression solves.

Other step and parameter are identical with one of embodiment one to nine.

Adopt following examples to verify beneficial effect of the present invention:

Embodiment mono-:

By adopting double-wavelength light source λ ₁=750nm, λ ₂=830nm, light source S is that light source detection device spacing is 10mm to the air line distance of detecting device D1, light source S is that light source detection device spacing is 35mm to the air line distance of detecting device D2.The near infrared light that this setting can make D2 detect can effectively penetrate cerebral cortex, and the near infrared light that D1 detects only penetrates an outer brain tissue.The absorbance of acquisition is changed to the time series △ [HbO that changes oxyhemoglobin concentration change amount by revising langbobier law into ₂] ⁿ(k), △ [HbO ₂] ^fand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ⁿ(k), △ [HHb] ^f(k).By empirical mode decomposition algorithm to near-end hemodynamics variation △ [HbO ₂] ⁿor △ [HHb] (k) ⁿ(k) be decomposed into solid-state mode function component (IMF).This IMF component has two dimension directions.The first dimension direction is the natural mode component exponent number that empirical mode decomposition algorithm decomposes, the time scale that this dimension direction is corresponding different, and the larger time scale through natural mode component of exponent number value is less; Another dimension direction is the moment of this IMF component, this dimension direction indication signal temporal evolution.Be that two-dimensional signal is carried out linear combination estimation △ [HbO by IMF component with component exponent number and time dimension ₂] ^for △ [HHb] (k) ^f(k) physiology in disturbs, and will build error signal s (k) by adaptive filter algorithm.Solve accumulative total square error performance function J (k) minimum that makes error signal s (k) by least-squares estimation criterion, s (k) rejects by auto adapted filtering the cerebration signal that physiology disturbs.The difference of the method is that the estimation that physiology disturbs utilizes two-dimensional signal to carry out combinational estimation, has not only considered exponent number dimension information but also considered time dimension information, and the Interference Estimation value of acquisition is more accurate.

Claims (10)

1. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method, is characterized in that: the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method specifically carries out according to following steps:

Step 1, the near-infrared probe that uses light source S and two detecting device D1 and D2 to form in brain tissue to be measured position are surveyed;

Step 2, detection diffuse reflection light intensity signal also convert the electric signal that obtains reaction intensity signal through photoelectric sensor;

Step 3, electric signal are obtained the time series △ [HbO of the oxyhemoglobin concentration change amount that detecting device D1 records by revising langbobier law ₂] ⁿand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ⁿ(k) the time series △ [HbO of the oxyhemoglobin concentration change amount that, detecting device D2 records ₂] ^fand the time series △ [HHb] of reduced hemoglobin concentration change amount (k) ^f(k);

Step 4, use x (k) represent △ [HbO ₂] ⁿor △ [HHb] (k) ⁿ(k);

Step 5, use y (k) represent △ [HbO ₂] ^for △ [HHb] (k) ^f(k);

In step 6, y (k), comprise cerebration signal r (k) and physiology and disturb i (k), i.e. y (k)=r (k)+i (k), physiology disturbs and can be expressed as

$\hat{i}\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=k-M}{\overset{k+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right);$

Step 7, according to y (k)=r (k)+i (k) and can extrapolate the expression formula of brain function signal;

Step 8, utilize weighted least square algorithm as cost function, ask for optimized coefficients w _i,j(k), then by the optimized coefficients w asking for _i,j(k) bring the expression formula of brain function signal into, can obtain brain function signal s (k).

2. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 1, it is characterized in that: in described step 1, light source S adopts integrated dual wavelength near-infrared light source, light source S is r to the air line distance between near-end detecting device D1 ₁; Light source S is r to the air line distance between remote sensor D2 ₂.

3. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 2, is characterized in that: two kinds of wavelength that dual wavelength near-infrared light source sends are respectively λ ₁=750nm, λ ₂=830nm.

4. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 3, is characterized in that: the spacing r of light source S and detecting device D1 ₁for 10mm, the spacing r of light emitting source S and detecting device D2 ₂for 35mm.

5. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 4, is characterized in that: △ [HbO in described step 3 ₂] ⁿand △ [HHb] (k) ⁿ(k) be:

$\mathrm{\&Delta;}{\left[{\mathrm{HbO}}_2\right]}^N\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^N\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^N\left(k\right)/\mathrm{DPF})}{r_1({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

$\mathrm{\&Delta;}{\left[\mathrm{HHb}\right]}^N\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^N\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{H{\mathrm{bO}}_2}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^N\left(k\right)/\mathrm{DPF})}{r_1({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

△ [HbO ₂] ^fand △ [HHb] (k) ^f(k) be:

$\mathrm{\&Delta;}{\left[{\mathrm{HbO}}_2\right]}^F\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^F\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^F\left(k\right)/\mathrm{DPF})}{r_2({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

$\mathrm{\&Delta;}{\left[\mathrm{HHb}\right]}^F\left(k\right)=\frac{({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_1}^F\left(k\right)/\mathrm{DPF})-({\mathrm{\&epsiv;}}_{H{\mathrm{bO}}_2}\left({\mathrm{\&lambda;}}_1\right)\mathrm{\&Delta;}{A}_{{\mathrm{\&lambda;}}_2}^F\left(k\right)/\mathrm{DPF})}{r_2({\mathrm{\&epsiv;}}_{\mathrm{Hb}{O}_2}\left({\mathrm{\&lambda;}}_2\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_1\right)-{\mathrm{\&epsiv;}}_{{\mathrm{HbO}}_2}\left({\mathrm{\&lambda;}}_1\right){\mathrm{\&epsiv;}}_{\mathrm{HHb}}\left({\mathrm{\&lambda;}}_2\right))}$

Wherein, ε _hHb(λ ₁) for optical source wavelength be λ ₁time the extinction coefficient of reduced hemoglobin, ε _hHb(λ ₂) for optical source wavelength be λ ₂time the extinction coefficient of reduced hemoglobin,

(λ ₁) for optical source wavelength be λ ₁time oxyhemoglobin extinction coefficient, (λ ₂) for optical source wavelength be λ ₂time oxyhemoglobin extinction coefficient,

with be illustrated in S-D1 measurement passage and wavelength and be λ respectively ₁and λ ₂time absorbance variable quantity time series, with be illustrated in S-D2 measurement passage and wavelength and be respectively λ ₁and λ ₂time absorbance variable quantity time series,

K is the time, k=1, and 2 ..., N; N is positive integer;

DPF is the differential path factor.

6. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 5, it is characterized in that: x in described step 4 (k) utilizes empirical mode decomposition algorithm that x (k) is decomposed into N solid-state mode function component IMF component, using residual components as last IMF component, x (k) is expressed as

$x\left(k\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{c}_i\left(k\right)$

Wherein, c _i(k) the IMF component for decomposing; N is positive integer.

7. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 6, it is characterized in that: y in described step 6 (k)=r (k)+i (k), for specific time point k, select the time window [k-M of the 2M+1 length centered by this time point, k-M+1,, k+M], utilize two-dimentional weight coefficient w _i,j(k) relation that IMF component and physiology disturb is described;

Described in step 6 wherein for the estimation of i (k), i=1,2 ..., N, w _i,j(k) be optimized coefficients, c _i(j) i empirical modal component of expression is in the size in j moment, and wherein, M is half window length, and k is positive integer, and N is positive integer.

8. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 7, is characterized in that: the expression formula of brain function signal in described step 7:

$S\left(k\right)=y\left(k\right)-\hat{i}\left(k\right)=r\left(k\right)+[i\left(k\right)-\hat{i}\left(k\right)]$

Wherein, s (k) is brain function signal, and r (k) is the estimation of brain function signal s (k), and i (k) is physiology interference, for the estimation of i (k).

9. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 8, it is characterized in that: in described step 8, utilize weighted least square algorithm, obtain brain function signal s (k), for:

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{s}^2\left(n\right)$

Further be expressed as

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]}^2$

Wherein, J (k) is square error performance function; χ is exponential weighting factor, χ=0.99, and n=1 ... k, k is positive integer, i=1,2 ..., N, N is positive integer, M is half window length, w _i,j(k) be i IMF component at the weight coefficient optimized coefficients in j moment, c _i(j) i empirical modal component of expression is at the size in j moment, s ²(n) be moment n brain function signal s (n) square, y (n) obtains brain function signal for moment n utilizes S-D2 passage, solves the w that makes J (k) minimum _i,j(k), obtain brain function signal s (k).

10. the brain-computer interface method for extracting signal based on empirical mode decomposition and time slip-window weighted least-squares method according to claim 9, is characterized in that: in described step 8, the preparation method of brain function signal s (k) is:

One, accumulative total square error performance function J (k) minimum that represents to make error brain function signal s (k) by weighted least square algorithm estimation criterion, J (k) is expressed as:

$J\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{s}^2\left(n\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]}^2$

W in formula _i,j(k) be optimized coefficients, c _i(j) i empirical modal component of expression is in the size in j moment, and y (n) obtains brain function signal for moment n utilizes S-D2 passage, n=1 ... k, k is positive integer, i=1,2 ..., N, N is positive integer, M is half window length;

Two, solve optimization coefficient w _i,j(k):

By to J (k) with respect to w _i,j(k) differentiate, and make it equal zero,

$-2\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}[y\left(n\right)-\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)c_i\left(j\right)]c_l\left(m\right)=0$

W in formula _i,j( ^k) be i empirical modal component at the weight coefficient in j moment, c _i(j) i empirical modal component of expression is at the size in j moment, c _l(m) for representing the size of l empirical modal component in the m moment

Obtained by above formula

$\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}y\left(n\right)c_l\left(m\right)=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=n-M}{\overset{n+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{{\mathrm{\&chi;}}^{k-n}c}_i(n-j)c_l(n-m)$

Or

$\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{j=k-M}{\overset{k+M}{\mathrm{\&Sigma;}}}{w}_{i,j}\left(k\right)R_{i,l;j,m}\left(k\right)=p_{l,m}\left(k\right),l=1,...,N,m=k-M,...,k+M$

In formula, k is positive integer, wherein, and p _l,mand R (k) _{i, l; j,m}(k) expression formula is

$p_{l,m}\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}y\left(n\right)c_l\left(m\right)$

$R_{i,l;j,m}\left(k\right)=\underset{n=1}{\overset{k}{\mathrm{\&Sigma;}}}{\mathrm{\&chi;}}^{k-n}{c}_i\left(j\right)c_l\left(m\right)$

In formula, k is positive integer, c _i(j) i empirical modal component of expression is at the size in j moment, c _l(m), for l empirical modal component of expression is in the size in m moment, y (n) obtains brain function signal for moment n utilizes S-D2 passage;

Being expressed as of its matrix form

Can further be reduced to

R(k)w(k)＝p(k)

If matrix R (k) is nonsingular, optimal coefficient calculates by following formula

w ^*(k)＝R ^-1(k)p(k)

Wherein, w ^*(k) be expressed as the optimum solution of w (k),

R ^-1(K) be the inverse matrix of R (K),

$w\left(k\right)=\left[\begin{array}{c}{w}_{1,-M}\left(k\right)\\ w_{1,-M+1}\left(k\right)\\ .\\ .\\ .\\ w_{N,M}\left(k\right)\end{array}\right],$

$p\left(k\right)=\left[\begin{array}{c}{p}_{1,-M}\left(k\right)\\ p_{1,-M+1}\left(k\right)\\ .\\ .\\ .\\ p_{N,M}\left(k\right)\end{array}\right],$

$c\left(k\right)=\left[\begin{array}{c}{c}_1(-M)\\ c_1(-M+1)\\ .\\ .\\ .\\ c_N\left(M\right)\end{array}\right];$

Three, solve brain function signal s (k):

s(k)＝y(k)-c ^T(k)w ^*(k)，

Wherein c ^t(k) what represent is the transposed matrix of c (k), w ^*(k) optimal coefficient vector that expression solves.