CN102657526B - Method for evaluating R value and power spectrums of electroencephalogram signals causing discomfort when people watch three-dimensional (3D) images - Google Patents

Abstract

Provided is a method for evaluating an R value and power spectrums of electroencephalogram signals causing discomfort when people watch three-dimensional (3D) images. The method comprises arranging an experimental environment and experimental conditions, arranging experimental sequence of a subject, explaining the experiment process, arranging an electroencephalogram electrical electrode, collecting electroencephalogram signals of the subject before watching images for 2 minutes, and performing a first questionnaire survey; collecting electroencephalogram signals when the subject watches the whole movie and performing a second questionnaire survey after the movie is over; collecting electroencephalogram signals for 5 minutes when the subject rests for 10-20 minutes after watching the movie and performing a third questionnaire survey; preprocessing the collected electroencephalogram signals; analyzing the power spectrums in a plurality of wave bands according to the frequency range obtained by preprocessing the electroencephalogram signals; and analyzing the R value and performing statistics t-inspection under 2D experimental conditions and 3D experimental conditions. The method can perform comparison of 2D images and the 3D images, analyzes centralized brain regions of the electroencephalogram signals from the energy angle, provides an experimental basis for follow-up researches, and then provides thoughts for technical standards for evaluating the 3D images.

Description

Watch the 3D image to cause EEG signals power spectrum and the evaluation methodology of R value of sense of discomfort

Technical field

The present invention relates to a kind of collection and feature extracting method of EEG signals.Particularly relate to EEG signals power spectrum and the evaluation methodology of R value that a kind of 3D of watching image causes sense of discomfort

Background technology

The 3D TV is the abbreviation of 3 D stereoscopic image TV.It is slightly variant that the 3 D stereoscopic image TV utilizes people's eyes to observe the angle of object, therefore can distinguish the object distance, produce three-dimensional this principle of vision, the image that right and left eyes is seen separates, thereby makes the user without by anaglyph spectacles, can bore hole experiencing stereoperception.The improvement to picture quality is not only in the generation of 3D TV, especially to change vivid in image.Although it is very strong that stereo display technique has advantages of third dimension and feeling of immersion, if but watch for a long time stereoscopic image (more than 30 minutes), just probably produce visual fatigue, the uncomfortable symptom such as dizzy, this has affected the universal and development of stereo display technique to a certain extent.Health and safety is also that a difficult problem of capturing is badly in need of in 3D stereo display.Result of study shows, spectators, when watching stereoscopic image, because eyes can promptly move around, thereby easily cause the sense of discomfort such as eyestrain.

The people such as Li estimate visual fatigue with background brain wave and event related potential, and result shows that the frequency spectrum of background brain wave and the preclinical delay of P700 all depend on visual disparity and viewing time, and this point also is verified in the measuring method of subjectivity.Therefore, the delay of electrical wave measurement's visual information in transmitting procedure of requiring mental skill is a kind of effective ways of estimating visual fatigue.For the senses of discomfort such as visual fatigue of watching 3-dimensional image to cause, corresponding research work is carried out also fewerly.

For a new technology product, freely watch and need to be guaranteed, therefore, picture quality and visual comfort all have advantage than conventional television.If can find a kind of three-dimensional stereo display technique, when can greatly reduce existing stereoscopic display device cost, can be simple and easy to again with, take up space limited, most importantly can alleviate greatly or solve numerous physiology and the psychological problem that present stereo display technique brings to people, the development that will inevitably promote stereo display technique, allow 3 D stereo show that science service is in popular life.

Summary of the invention

Technical problem to be solved by this invention is, the similarities and differences of human body electroencephalogram's signal are provided while providing a kind of basis to watch 2D, 3D image, thereby a kind of method of assessing visual fatigue is provided, promotes the 3D image of watching of 3D image technology development to cause EEG signals power spectrum and the evaluation methodology of R value of sense of discomfort.

The technical solution adopted in the present invention is: a kind of 3D of watching image causes EEG signals power spectrum and the evaluation methodology of R value of sense of discomfort, comprises the steps:

(1) experimental situation and condition are set, arrange experimenter's experimental sequence;

(2) explain orally experiment flow, settle electrode for encephalograms, gather the front EEG signals 2min of experimenter's viewing, and carry out questionnaire survey for the first time, described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(3) watch whole film and gather EEG signals, carrying out questionnaire survey for the second time after wrapped film, described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(4) experimenter, after finishing the viewing stage, after the 10～20min that has a rest, gathers EEG signals 5min, carries out questionnaire survey for the third time, and described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(5) gathered EEG signals is carried out to pretreatment, comprising:

(a) EEG signals gathered is reseted to reference, filtering de-noising and normalized;

(b) utilize the ICA method to remove the eye electricity, be about to signal decomposition and become each independently component component is analyzed;

(6), for the frequency range of pretreatment gained EEG signals, carry out the power spectrumanalysis of a plurality of wave bands:

To the experimental data collected, on MATLAB 7.11 platforms, processed; At first, eeg data is averaged to power spectrum to be solved, choose corresponding data point and carry out Fourier transformation, then get the product of frequency spectrum and this frequency spectrum conjugation, obtain the estimation of power spectrum, calculate the percent value that each wave band accounts for general power spectrum, and to whole experiment duration totally 85～100min carry out fitting a straight line, seek overall variation trend; Secondly, on the SPSS18.0 platform, respectively power spectrum numerical value and percent value are carried out to one factor analysis of variance, in order to judge the difference under two kinds of experiment conditions of 2D and 3D;

(7) the R value is analyzed, and carries out the statistics t-check of two kinds of experiment condition 2D and 3D;

Step 1) comprising: the bias light illumination arranged in room is 2.5～3.5Lux, select the experimenter to watch for the first time the film that will show, the 3D television set is positioned at the dead ahead of experimenter 2.5～3.5m, when watching the 3D image, wear the active shutter type 3 D spectacles of 120/second of field frequencies, whole experimental period is 85～100min.

Step 1) also comprise: the experimenter is divided into to two groups, first watches the 3D image mode for one group, after watch the 2D image mode, another group is first watched the 2D pattern, after watch the 3D pattern.

Step 2) the front EEG signals 2min of described collection viewing, be to gather the experimenter to test the EEG signals under front quiescent condition, the 1min that closes one's eyes, and the 1min that opens eyes, the sampling of EEG signals guarantees that electrode impedance is less than 5000 ohm.

Step 3) describedly watch each experimenter of film requirement to watch the interval of two images experiments of 2D and 3D will be more than 10 days.

Step 5) the described ICA method of utilizing is removed the eye electricity, specifically:

In the situation that unknown source signal S and this unknown source signal S aliasing characteristic A only extract source signal from the mixed signal X=AS received, that is: one group of source signal s (t)=[s independently ₁(t) ..., s _n(t)] ^TMix through hybrid system A, obtain observation signal x (t)=[x ₁(t) ..., x _n(t)] ^T, be expressed as follows:

$\left(\begin{array}{c}{x}_1\left(t\right)\\ x_2\left(t\right)\\ \·\\ \·\\ \·\\ x_n\left(t\right)\end{array}\right)=A\left(\begin{array}{c}{s}_1\left(t\right)\\ s_2\left(t\right)\\ \·\\ \·\\ \·\\ s_n\left(t\right)\end{array}\right)$

Wherein, source signal s (t) and hybrid system A are unknown, only have mixed x (t) to observe, and the task of ICA is under the prerequisite of A and S the unknown, find and separate mixed matrix W, make output matrix

U＝W·X＝W·A·S

Wherein, x (t)=[x ₁(t) ..., x _n(t)] ^TFor observation signal, s (t)=[s ₁(t) ..., s _n(t)] ^TFor source signal, A is hybrid system, and S is source signal, and U is output matrix, and W is for separating mixed matrix, and X is mixed signal

Step 7) the average power spectra ratio that uses alpha to be two frequency bands of 18-22Hz for 10-13Hz and beta causes an index of sense of discomfort as the 3D image, establish signal and in the mean power spectrum density of frequency band h be:

$G\left(h\right)=\frac{{\∫}_{f_d}^{f_u}p\left(f\right)\mathrm{df}}{f_u-f_d}$

$R=\frac{G\left(h_1\right)}{G\left(h_2\right)}$

Wherein, G is the mean power spectrum density, the frequency band that h is signal, f _uFor the upper limit of frequency band h, f _dFor the lower limit of frequency band h, the power spectral density that p (f) is signal, h ₁And h ₂For the different frequency bands of EEG signals, the ratio that R is the result of calculation definition;

All experimenters are carried out to the calculating of R value, and the R value of 0～10min and 35～45min is carried out to the t-check.

The 3D of watching image of the present invention causes EEG signals power spectrum and the evaluation methodology of R value of sense of discomfort, can carry out the contrast of 2D and 3D image, analyze the collection Midbrain Area of EEG signals from energy point of view, for follow-up research provides experiment basis, thereby, for the technical standard of estimating the 3D image provides thinking, promote the development of 3D image technology.

The accompanying drawing explanation

Fig. 1 is the experiment figure that leads that adopts in the present invention;

Fig. 2 is the flow chart of the inventive method;

Fig. 3 is the lead average power spectra ratio figure of delta wave band of FP2;

Fig. 4 is under 2D and two kinds of experiment conditions of 3D, and Oz leads and locates the average power spectra of alpha wave band;

Fig. 5 (a) is the power spectrum figure of the gamma wave band of 2D experimental group;

Fig. 5 (b) is the power spectrum figure of the gamma wave band of 3D experimental group;

Fig. 6 is that under the 3D condition, the R value after 0-10min and 35-45min is t-assay figure.

The specific embodiment

The EEG signals power spectrum and the evaluation methodology of R value that the 3D of watching image of the present invention are caused to sense of discomfort below in conjunction with embodiment and accompanying drawing are described in detail.

The 3D of watching image of the present invention causes EEG signals power spectrum and the evaluation methodology of R value of sense of discomfort, as shown in Figure 2, comprises the steps:

(1) experimental situation and condition are set, arrange experimenter's experimental sequence;

Comprise: the bias light illumination arranged in room is 2.5～3.5Lux, select the experimenter to watch for the first time the image that will show, the 3D television set is positioned at the dead ahead of experimenter 2.5～3.5m, when watching the 3D image, wear the active shutter type 3 D spectacles of 120/second of field frequencies, whole experimental period is 85～100min.

Also comprise: the experimenter is divided into to two groups, first watches the 3D pattern for one group, after watch the 2D pattern, another group is first watched the 2D pattern, after watch the 3D pattern.

The 64 conducts digital eeg recording instrument that the present embodiment adopts U.S. Neuroscan company to produce gather eeg data.The experimenter is arranged in a room that electromagnetic shielding is good, soundproof effect is good, temperature humidity is suitable and is tested.Block external light source by lighttight curtain, the bias light illumination in room is 3Lux.The experimenter to feel comfortable but the posture that does not affect data acquisition sit in an armchair.Dead ahead apart from experimental subject 3m left and right is 55 inches 3D TVs of a Samsung 55C8000XF, and the image of watching is the 3D animation: " Monster Aliens " and " Megamind ".When watching the 3D image, the experimenter wears the active shutter type 3 D spectacles of 120/second of field frequencies.Whole experimental period is the 95min left and right.In this experiment, the experimenter is 10 people, and 5 people first watch the 3D pattern, after watch the 2D pattern, 5 people first watch the 2D pattern, after watch the 3D pattern.

(2) explain orally experiment flow, settle electrode for encephalograms, gather the front EEG signals 2min of experimenter's viewing, and carry out questionnaire survey for the first time, described questionnaire comprises the problem of the aspects such as visual discomfort sense, body part discomfort;

EEG signals 2min before described collection viewing, be to gather the experimenter to test the EEG signals under front quiescent condition, the 1min that closes one's eyes, and the 1min that opens eyes, the sampling of EEG signals guarantees that electrode impedance is less than 5000 ohm.

This period is the front preparatory stage of experiment, and this period is introduced experiment flow by the experimenter to the experimenter.

The placement location of electrode adopts international 10/20 system standard, as shown in Figure 1.

Electrode adopts the Ag/AgCl electrode, and usings left ear-lobe (M1) and auris dextra and hang down (M2) as bipolar reference, and the sample frequency of EEG signals is 1000Hz, and adopts the 50Hz wave trap to remove the power frequency interference.In experiment, guarantee that electrode impedance is less than 5000 ohm.

(3) watch whole image and gather EEG signals, image carries out questionnaire survey for the second time after finishing, and described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

The described interval of watching each experimenter of image requirement to watch 2D and two images of 3D to test will be more than 10 days.

This period is the viewing period, and this window duration of the present embodiment 80min, require experimenter's peace and quiet to be sitting on armchair, watches image.According to the experiment progress, watch respectively the image of 2D or 3D pattern, when image finishes, carry out questionnaire survey for the second time.At whole viewing experimental session, the experimenter watches image for the first time, does not have priori, in experimentation, requires the experimenter to keep relaxation state, does not allow any actual act.Whole experimental program requires each experimenter to complete 2D and two kinds of experiments of 3D, between two kinds of experiments, guarantees that each experimenter interval is more than 10 days, in order to carry out fatigue recovery.

(4) experimenter, after finishing the viewing stage, after the 10～20min that has a rest, gathers EEG signals 5min, carries out questionnaire survey for the third time, and described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

This period is convalescent period, and in the present embodiment, the experimenter is after finishing the viewing stage, and rest 15min, carry out questionnaire survey again, and gathers the eeg data after having a rest, duration 5min.

(5) gathered EEG signals is carried out to pretreatment, comprising:

(a) EEG signals gathered is reseted to reference, filtering de-noising and normalized;

After data acquisition, be the accuracy that guarantees that later data is analyzed, adopted and reseted reference, filtering de-noising and normalized.

(b) utilize the ICA method to remove the eye electricity, be about to signal decomposition and become each independently component component is analyzed;

Independent component analysis (ICA) is decomposed signal, is broken down into each independently component component is analyzed.In the situation that unknown source signal S and this unknown source signal S aliasing characteristic A only extract source signal from the mixed signal X=AS received.That is: one group of source signal s (t)=[s independently ₁(t) ..., s _n(t)] ^TMix through linear system A, obtain observation signal x (t)=[x ₁(t) ..., x _n(t)] ^T.Be expressed as follows:

$\left(\begin{array}{c}{x}_1\left(t\right)\\ x_2\left(t\right)\\ \·\\ \·\\ \·\\ x_n\left(t\right)\end{array}\right)=A\left(\begin{array}{c}{s}_1\left(t\right)\\ s_2\left(t\right)\\ \·\\ \·\\ \·\\ s_n\left(t\right)\end{array}\right)---\left(1\right)$

Wherein, source signal s (t) and hybrid system A are unknown, only have mixed x (t) to observe.The task of ICA is under the prerequisite of A and S the unknown, finds and separates mixed matrix W, makes output matrix

U＝W·X＝W·A·S (2)

In formula, x (t)=[x ₁(t) ..., x _n(t)] ^TFor observation signal, s (t)=[s ₁(t) ..., s _n(t)] ^TFor source signal, A is hybrid system, and S is source signal, and U is output matrix, and W is for separating mixed matrix, and X is mixed signal

The EEG signals gathered in experiment, because repeatedly blinking, obviously affect the calculating of average power spectra value.We use the ICA method to separate pretreated EEG signals, find component and the noise component(s) the strongest with the signal of blinking dependency, remove this two components, data after the mixed matrix W of the solution of having preserved and removal component are inverted, thereby reconstruct the EEG signals of removing the eye electricity, for the follow-up analysis to energy lays the first stone.

(6), for the frequency range of pretreatment gained EEG signals, carry out the power spectrumanalysis of a plurality of wave bands:

To the experimental data collected, on MATLAB 7.11 platforms, processed; At first, eeg data is averaged to power spectrum to be solved, choose corresponding data point and carry out Fourier transformation, then get the product of frequency spectrum and this frequency spectrum conjugation, obtain the estimation of power spectrum, calculate the percent value that each wave band accounts for general power spectrum, and to whole experiment duration totally 85～100min carry out fitting a straight line, seek overall variation trend; Secondly, on the SPSS18.0 platform, respectively power spectrum numerical value and percent value are carried out to one factor analysis of variance, in order to judge the difference under two kinds of experiment conditions of 3D and 2D;

Because duration of experiment is longer, the method adopted in the present embodiment is that signal is carried out to segment processing, get the 10s data every 20min, alpha (8-13Hz), beta (13-30Hz), delta (1-4Hz), four wave bands of theta (4-8Hz) are carried out to power spectrumanalysis, from the power spectrumanalysis result, subjects's sense of discomfort is apparent in view after about 50-60min, and concrete outcome is as shown in Fig. 3 to Fig. 5.

Shown in the viewing process result that record data are analyzed of leading to FP2 in Fig. 3, the average power spectra ratio led at this from each wave band can be seen, FP2 is led after the data characteristics amount carries out fitting a straight line, and the power spectrum of delta wave band shows that 3D image experimental group presents ascendant trend; 2D image experimental group presents rising trend.Extremely tired and lethargy occurs because the delta wave band shows the adult, this presentation of results is compared to watches the 2D image, watches the experimenter of 3D image to reveal more obvious sense of discomfort at frontal region brain ammeter.

For power spectrum situation over time, with the average profile diagram, compare.Figure 4 shows that EEG signals segmentation in every 10 minutes to recording in experimentation calculates the mean power spectral density value, carry out the variance analysis of repeated measure, the coordinate points of abscissa time represents respectively 0,10,20,30,40,50,60,70,80min.In the calculating that averages power spectrum, the alpha wave band that Oz leads and locates, after viewing starts, increase trend all appears in the situation of the image of 2D and two kinds of patterns of 3D, and the growth rate of 3D image is apparently higher than the 2D image.

Gamma wave band to two experimental grouies is contrasted, and as seen from Figure 5,2 FPz that lead of frontal lobe, FP2 wave form varies trend are consistent, and 3 POz that lead, Oz, the O1 wave form varies trend of occipital lobe are consistent; 3D image experimental group changes more obvious than 2D image experimental group, after viewing 50min, reach peak value.

The people is under the situation felt the stress, high band EEG signals (as the gamma ripple) is easy to observe, in Fig. 5 (a), be the power spectrum figure of the gamma wave band of 2D image experimental group, Fig. 5 (b) is the power spectrum figure of the gamma wave band of 3D image experimental group, gamma wave band analysis result is shown after the long period viewing to the tested situation that psychentonia occurs, feel the stress and increase.

(7) the R value is analyzed, and carries out the statistics t-check of two kinds of experiment condition 2D and 3D;

The average power spectra ratio of use alpha (10-13Hz) and two frequency bands of beta (18-22Hz) causes an index of sense of discomfort as the 3D image, establish signal and in the mean power spectrum density of frequency band h be:

$G\left(h\right)=\frac{{\∫}_{f_d}^{f_u}p\left(f\right)\mathrm{df}}{f_u-f_d}---\left(3\right)$

$R=\frac{G\left(h_1\right)}{G\left(h_2\right)}---\left(4\right)$

Wherein, G is the mean power spectrum density, the frequency band that h is signal, f _uFor the upper limit of frequency band h, f _dFor the lower limit of frequency band h, the power spectral density that p (f) is signal, h ₁And h ₂For the different frequency bands of EEG signals, the ratio that R is the result of calculation definition.

10 experimenters are carried out to the calculating of R value, and the R value of 0～10min and 35～45min is carried out to the t-check, as shown in Figure 6, wherein, for simplifying the analysis, the present embodiment has only adopted 21 in the standard lead to lead to result.After viewing 30min, the notable difference of R value has appearred in frontal region (leading for No. 1-3).

Final result shows, in the situation that other experiment conditions are identical, with watching 3D image time lengthening, easily causes that the subjects produces fatigue state.Watch the difference of 2D and two kinds of EEG signals that image causes of 3D, comparatively obvious with the reaction in occipital lobe and frontal lobe Liang Genao district.EEG signals power spectrum and the evaluation methodology of R value that the 3D image causes sense of discomfort can be used as a research direction further investigated, for the evaluation index that the 3D image further is discussed has been established strong basis.

Claims (6)

1. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image, is characterized in that, comprises the steps:

(1) experimental situation and condition are set, arrange experimenter's experimental sequence;

(2) explain orally experiment flow, settle electrode for encephalograms, gather the front EEG signals 2min of experimenter's viewing, and carry out questionnaire survey for the first time, described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(3) watch whole film and gather EEG signals, carrying out questionnaire survey for the second time after wrapped film, described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(4) experimenter, after finishing the viewing stage, after the 10～20min that has a rest, gathers EEG signals 5min, carries out questionnaire survey for the third time, and described questionnaire comprises the problem of visual discomfort sense, body part discomfort;

(5) gathered EEG signals is carried out to pretreatment, comprising:

(a) EEG signals gathered is reseted to reference, filtering de-noising and normalized;

(b) utilize the ICA method to remove the eye electricity, be about to signal decomposition and become each independently component component is analyzed;

(6), for the frequency range of pretreatment gained EEG signals, carry out the power spectrumanalysis of a plurality of wave bands:

To the experimental data collected, on the MATLAB7.11 platform, processed; At first, eeg data is averaged to power spectrum to be solved, choose corresponding data point and carry out Fourier transformation, then get the product of frequency spectrum and this frequency spectrum conjugation, obtain the estimation of power spectrum, calculate the percent value that each wave band accounts for general power spectrum, and to whole experiment duration totally 85～100min carry out fitting a straight line, seek overall variation trend; Secondly, on the SPSS18.0 platform, respectively power spectrum numerical value and percent value are carried out to one factor analysis of variance, in order to judge the difference under two kinds of experiment conditions of 2D and 3D;

(7) the R value is analyzed, and carries out the statistics t-check of two kinds of experiment condition 2D and 3D;

Use alpha for the index of 10-13Hz and the beta average power spectra ratio that is two frequency bands of 18-22Hz as sense of discomfort due to the 3D image, establish signal and in the mean power spectrum density of frequency band h be:

$G\left(h\right)=\frac{{\∫}_{f_d}^{f_u}p\left(f\right)\mathrm{df}}{f_u-f_s}$

$R=\frac{G\left(h_1\right)}{G\left(h_2\right)}$

Wherein, G is the mean power spectrum density, the frequency band that h is signal, f _uFor the upper limit of frequency band h, f _dFor the lower limit of frequency band h, the power spectral density that p (f) is signal, h ₁And h ₂For the different frequency bands of EEG signals, the ratio that R is the result of calculation definition;

All experimenters are carried out to the calculating of R value, and the R value of 0～10min and 35～45min is carried out to the t-check.

2. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image according to claim 1, it is characterized in that, step (1) comprising: the bias light illumination arranged in room is 2.5～3.5Lux, select the experimenter to watch for the first time the film that will show, the 3D television set is positioned at the dead ahead of experimenter 2.5～3.5m, when watching the 3D image, wear the active shutter type 3 D spectacles of 120/second of field frequencies, whole experimental period is 85～100min.

3. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image according to claim 1, it is characterized in that, step (1) also comprises: the experimenter is divided into to two groups, first watch the 3D image mode for one group, after watch the 2D image mode, another group is first watched the 2D pattern, after watch the 3D pattern.

4. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image according to claim 1, it is characterized in that, EEG signals 2min before the described collection viewing of step (2), to gather the experimenter to test the EEG signals under front quiescent condition, 1min closes one's eyes, the 1min that opens eyes, the sampling of EEG signals guarantees that electrode impedance is less than 5000 ohm.

5. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image according to claim 1, it is characterized in that, the described interval of watching each experimenter of film requirement to watch 2D and two images of 3D to test of step (3) will be more than 10 days.

6. EEG signals power spectrum and the evaluation methodology of R value of watching sense of discomfort due to the 3D image according to claim 1, is characterized in that, the described ICA method of utilizing of step (5) is removed the eye electricity, specifically:

In the situation that unknown source signal S and this unknown source signal S aliasing characteristic A only extract source signal from the mixed signal X=AS received, that is: one group of source signal s (t)=[s independently ₁(t) ..., s _n(t)] ^TMix through hybrid system A, obtain observation signal x (t)=[x ₁(t) ..., x _n(t)] ^T, be expressed as follows:

$\left(\begin{array}{c}{x}_1\left(t\right)\\ x_2\left(t\right)\\ .\\ .\\ .\\ x_n\left(t\right)\end{array}\right)=A\left(\begin{array}{c}{s}_1\left(t\right)\\ s_2\left(t\right)\\ .\\ .\\ .\\ s_n\left(t\right)\end{array}\right)$

Wherein, source signal s (t) and hybrid system A are unknown, only have mixed x (t) to observe, and the task of ICA is under the prerequisite of A and S the unknown, find and separate mixed matrix W, make output matrix

U=W·X=W·A·S

Wherein, x (t)=[x ₁(t) ..., x _n(t)] ^TFor observation signal, s (t)=[s ₁(t) ..., s _n(t)] ^TFor source signal, A is hybrid system, and S is source signal, and U is output matrix, and W is for separating mixed matrix, and X is mixed signal.

Images