Improved EMD interference envelope extraction method based on extrapolation interpolation

Abstract

The invention relates to the field of signal processing, in particular to an improved EMD interference envelope extraction method based on extrapolation interpolation. The method comprises the following steps: inputting an interference image B to be processed, and extracting interference information B3 [sigma] of an emission / absorption line according to a 3 [sigma] rule; b-spline fitting is carried out on the B3 sigma by using the low-pass filtering characteristic of B-spline fitting, and normalization is carried out to obtain Bnor; and obtaining an interference envelope BL by using an improved EMD method. According to the method, the problem of sub-pixel Doppler frequency shift can be effectively solved, effective information can be effectively extracted under the condition of low contrast, meanwhile, transmission and accumulation of decomposition errors caused by an end point effect are effectively improved, and finally high-precision radial velocity inversion is achieved. And a theoretical basis is provided for researching planet formation and galaxy evolution and exploring life origin and extraterrestrial civilization.

Description

Technical field:

[0001] The invention relates to the field of signal processing, and in particular to an improved EMD interference envelope extraction method based on extrapolation and interpolation. Technical background:

[0002] Spectral imaging technology can simultaneously obtain imaging information and spectral information of the target under test, which can not only reflect the shape characteristics of the target, but also characterize the physical and chemical properties of the target. As a new type of spectral imaging technology, coherent dispersion spectral imaging technology combines an interferometer with a fixed delay and a medium- and low-resolution spectrometer. It can not only improve the problem of mutual limitation between spectral resolution and light flux in the dispersion type spectrometer, but also solve the problem of limited detection accuracy caused by moving parts in the traditional interference type spectrometer. This is of great significance for detecting targets with long distance, low brightness, and sensitivity to external disturbances, especially in the detection of exoplanets based on the radial velocity method. It has very promising prospects. When detecting exoplanets based on the coherent dispersion spectrometer, the frequency shift generated by the Doppler effect is converted into the phase change of the interference fringes through the coherent dispersion technology, and then the radial velocity information is obtained according to the phase change. The accurate measurement of the radial velocity is not only of great scientific significance for studying planet formation, galaxy evolution, and exploring the origin of life and extraterrestrial civilizations, but also can provide key parameters for special military activities. Therefore, how to improve the detection accuracy is the primary problem that needs to be solved at present.

[0003] In order to improve the accuracy of radial velocity detection, domestic and foreign scholars have improved the stability and energy utilization of the system through thermal compensation, vacuum tanks, constant temperature and pressure transformation, delayed pair cross cancellation and new interference technology, and have achieved corresponding results. The asymmetric common optical path coherent dispersion spectral imaging technology proposed in recent years has great potential in the field of exoplanet detection. Based on the traditional coherent dispersion technology, this technology replaces the Michelson interferometer with an asymmetric Sagnac interferometer, so that the two beams output from the interferometer can be effectively used by the subsequent optical path, improving the energy utilization of the system. In addition, the two interference beams in the Sagnac interferometer have the characteristics of a common optical path, which effectively improves the stability of the system, and has been verified from the principle and experiment.

[0004] However, in the process of using the coherent dispersion system to detect based on the line-of-sight velocity method, since the position of the detector and the detection band range will not change, the Doppler frequency shift information can only be measured through absorption or emission lines. In practice, the resolution of the detector is limited, and it is very difficult to accurately extract the position of the center wavelength of the absorption and emission lines. In addition, the frequency shift caused by the Doppler effect is very small, and the problem of sub-pixel frequency shift will occur, which seriously limits the detection accuracy. Moreover, the early research on asymmetric common optical path coherent dispersion technology was mostly focused on principle verification and upgrading of system hardware. There is relatively little discussion on the data processing work of line-of-sight velocity inversion of interference fringes obtained using asymmetric common optical path coherent dispersion technology, and most of the simulations and discussions are carried out under the ideal condition that the intensity distribution of the detection target is uniform. However, since the intensity distribution of the target light source is not uniform in reality, the variation of the interference fringe intensity with the optical path difference obtained after passing through the coherent dispersion system is not an ideal sine and cosine function, but is modulated by an envelope function. The influence of the envelope function on the accuracy of the line-of-sight velocity inversion includes the following points: 1. The interference envelope will cause the extreme points of the interference fringes to be submerged, making it difficult to extract effective interference information; 2. The existence of endpoint effects in the traditional EMD method will cause large fluctuations at both ends of the signal during envelope fitting, affecting the quality of modal decomposition, and these fluctuations will gradually propagate from the endpoints to the internal data, and in severe cases will affect the entire data sequence. Summary of the invention:

[0005] The present invention proposes an improved EMD interference envelope extraction method based on extrapolation interpolation to solve the problems existing in the prior art of the extreme points of interference fringes being submerged due to the interference envelope, the endpoint effect in the traditional EMD method, and the difficulty in extracting effective interference information.

[0006] In order to achieve the purpose of the present invention, the technical solution provided by the present invention is: an improved EMD interference envelope extraction method based on extrapolation interpolation, comprising the following steps:

[0007] Step 1: Input the interference image B to be processed and extract the interference information B of the absorption / emission line according to the 3σ rule 3σ ;

[0008] Step 2: Use the low-pass filter characteristics of B-spline fitting to 3σ Perform B-spline fitting and normalize to get B nor；

[0009] Step 3: Obtain the interference envelope BL using the improved EMD method. The specific steps of the improved EMD method are:

[0010] (I) Use the differential method to find the B that is submerged due to envelope modulation nor The maximum and minimum points of ;

[0011] (ii) judging whether the endpoint signal is an extreme value according to the periodic characteristics of the interference fringes, and if it is an extreme value, judging whether it is a maximum value or a minimum value, and proceeding to step (iii); if it is not an extreme value, after performing endpoint extension, judging whether it is a maximum value or a minimum value, and proceeding to step (iii);

[0012] (iii) fitting the maximum (upper) and minimum (lower) envelope curves by cubic spline interpolation, and solving the average value M of the extreme envelope curves;

[0013] (IV) Using the original function B nor Subtract the mean value M of the envelope to get B′;

[0014] (V) Use Rilling stop condition to judge:

[0015] When the stop condition is met, the output envelope BL is output, BL = B nor -B′;

[0016] If the stopping condition is not met, let B nor =B′, continue to repeat the above steps (i) to (v) until the stop condition is met.

[0017] Furthermore, in the above step 1, the wave number width of the extracted absorption / emission line Δk≥2×(ceil(3σ / k int )+1)×k int , where k int is the detection wavenumber range of a single pixel, where ceil(3σ / k int ) represents the conversion of 3σ / k int Round to the nearest integer greater than or equal to this value (3σ / k int is an integer).

[0018] Furthermore, in the above step three (a), the discrete data is first fitted with B-spline, and the fitting process can filter out abnormal points such as interference glitches and intermittent noise in the original signal; then the second-order differential is calculated for the fitting result, and the minimum point of the second-order differential is used to represent the maximum point of the original interference signal, and the maximum point of the second-order differential represents the minimum point of the interference signal.

[0019] Furthermore, in the above step 3 (ii), the specific steps of performing endpoint extension are: for determining the extension range, first calculate the spacing between each extreme value, then based on the periodicity of the interference fringes and the symmetry of the light intensity envelope, combined with the spacing characteristics between the extreme points, use the ratio relationship of the extreme value spacing to determine the spacing between the extreme points of the original signal and the extended extreme point, and finally determine the extension range based on the distance between the last (first) extreme point of the original signal and the endpoint;

[0020] For the determination of the extended extreme value, when there is an inflection point near the endpoint, the function value at the inflection point is used for calculation according to the periodicity of the interference fringes and the symmetry of the light intensity envelope; when there is an extreme value near the endpoint, the ratio of the maximum (minimum) value to the minimum (large) value is first calculated, and then the ratio is extrapolated and interpolated, and finally the extended extreme value information is determined by combining the extrapolation and interpolation results with the extreme value information at both ends of the signal.

[0021] Compared with the prior art, the advantages of the present invention are:

[0022] 1. Based on the 3σ rule, the present invention uses a multi-column (row) superposition method to extract the interference information of the absorption or emission line required for inversion. The coherent dispersion system inverts the line-of-sight velocity information based on the phase change of the interference fringes, and the phase change amount Therefore, when the Doppler effect occurs, the wavelength on the detection unit will have a Δk shift, but when the coherent dispersion system parameters are fixed, the wave number (wavelength) range covered by each detector unit will not change. When the wave number range covered by the detector unit is very small and the frequency shift of the central wavelength of the absorption line or emission line is not equal to an integer multiple of the wavelength covered by a single pixel, the phenomenon of sub-pixel frequency shift will occur, so it is difficult to accurately obtain the phase shift caused by the Doppler frequency shift. If the wavelength range covered by a single pixel of the detector can contain most of the energy of the absorption / emission line, that is, the wavelength range of a single pixel is approximately equal to the wavelength range covered by the absorption / emission line, then the single pixel can still cover most of the energy of the absorption / emission line before and after the Doppler frequency shift, which can effectively solve the problem of sub-pixel frequency shift. In order to minimize the influence of background noise, the present invention takes the wavelength range within 3σ (including 99.7% of the energy) from the central wavelength of the absorption / emission line as the sampling of the detector; however, the large wavelength coverage range (large sampling) on ​​a single pixel will increase the resolution error of the line-of-sight velocity, so it is necessary to sample as small as possible to reduce the resolution range of the line-of-sight velocity. In order to balance the contradiction between using large sampling to solve the sub-pixel Doppler frequency shift and using small sampling to reduce the line-of-sight velocity resolution. The emission or absorption line is extracted by using small sampling and superposition of multiple columns in the dispersion direction. The sampling is designed according to the detection band and the resolution of the detector. In order to improve the line-of-sight velocity resolution accuracy and achieve the smallest possible sampling, all pixels in the effective area of ​​the detector are generally used, that is, the sampling is the ratio of the detection band to the row (column) resolution of the detector.

[0023] 2. The present invention can realize the effective extraction of extreme value information under low contrast conditions. When the extreme value of the interference envelope is relatively large relative to the interference fringe, the extreme value point of the fringe will be submerged, which is very unfavorable for obtaining the envelope by fitting the extreme value point in the EMD method, and will seriously affect the extraction accuracy of the interference envelope. For the interference envelope, the interference fringe belongs to the high-frequency component in the interference signal, and the high-frequency component can be enhanced in the form of differentiation. According to this property of the differential, the present invention combines the relationship between the second-order differential and the concavity of the signal, that is, when the second-order differential is greater than zero, the signal curve presents a concave property, and conversely, when the second-order differential is less than zero, the signal presents a convex property, and then judges the extreme value information of the original interference signal according to the extreme value of the second-order differential of the obtained interference signal. This is because the second-order differential reflects the rate of change of the slope of the interference signal, and the extreme value point of the second-order differential is the point where the slope changes the fastest, corresponding to the most convex or concave position of the curve. In the present invention, the most convex position of the curve is set as the maximum point of the interference fringe, and the most concave position is set as the minimum point, that is, the maximum value of the second-order differential corresponds to the minimum value of the original signal, and the minimum value of the second-order differential corresponds to the maximum value of the original signal.

[0024] 3. The present invention can effectively suppress the endpoint effect: When the traditional EMD method uses extreme value information to fit the upper and lower envelopes based on the cubic spline function, since the values ​​at the endpoints of the interference fringes are not necessarily extreme points, and there is a lack of any conditions nearby that can constrain them, the envelope obtained by fitting at the boundary will be different from the true envelope of the original signal, and the error will gradually accumulate during the iteration process. To solve this problem, the present invention combines the periodic characteristics of the interference fringes themselves and the symmetry of the light intensity envelope of a general light source, and performs endpoint extension design through the relationship between extreme value information. First, find the extreme value point of the original interference signal, and then determine whether the endpoint signal is an extreme value based on the periodic characteristics of the interference fringes. If it is an extreme value, then determine whether it is a maximum value or a minimum value. If it is not an extreme value, then it is necessary to perform endpoint extension. The present invention uses the extreme value information obtained by extension and the extreme value information of the original signal to extract the envelope mean, which can effectively improve the endpoint effect.

[0025] 4. According to the characteristics of interference fringes and the advantages of empirical mode decomposition (EMD) in processing nonlinear and non-stationary signals, such as completeness, filtering and adaptability, the present invention proposes an improved EMD interference envelope extraction method based on extrapolation and interpolation. It can not only realize the effective extraction of extreme value information of interference signals under low contrast, but also effectively improve the transmission and accumulation of decomposition errors caused by endpoint effects, and finally realize high-precision radial velocity inversion. Description of the drawings:

[0026] Figure 1 is the interference envelope extraction flow chart of the present invention;

[0027] Figure 2 It is the extreme value information diagram of the present invention, (a) absorption line, (b) emission line;

[0028] Figure 3 The method of the present invention is used to Figure 2 The envelopes extracted in (a) and (b) are shown in Figure 1.

[0029] Figure 4 The interference information and the visual velocity inversion result diagram after removing the envelope of the present invention, (a) absorption line, (b) emission line;

[0030] Figure 5 Figure 1 shows the interference information and visual velocity information before de-envelopment of the present invention (a) absorption line, (b) emission line. Specific implementation method:

[0032] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0033] The design idea of ​​the present invention is: firstly, based on small sampling, the interference information of the emission line or absorption line required for inversion is extracted using the 3σ rule, which solves the problem of sub-pixel Doppler frequency shift caused by the limited resolution of the detector. Then, the interference fringe information is enhanced by the differential method, which solves the problem of the difficulty in obtaining extreme values ​​under low contrast. Finally, combining the periodic characteristics of the interference fringes and the symmetry of the general light source intensity distribution, the extreme value information at the endpoint is obtained by extrapolation, interpolation and extension, which effectively improves the endpoint effect in the traditional EMD method.

[0034] See also Figure 1 The present invention provides an improved EMD interference envelope extraction method based on extrapolation interpolation, and the specific implementation steps are as follows:

[0035] Step 1: Input the interference image B to be processed and extract the interference information B of the absorption / emission line according to the 3σ rule 3σ . B 3σ It is the result of superposition of a certain number of columns (rows) of interference information along the dispersion direction on the detector. The number of columns (rows) required for superposition here is the result of dividing the required wavenumber width calculated according to the 3σ rule by the detection wavenumber width of a single pixel.

[0036] According to the basic principle of coherent dispersion technology, the half-width of the absorption line or emission line will increase after passing through the grating. Assuming that the half-width of the absorption / emission line is Δk n , the full width at half maximum after passing through the grating is due to the line spread function of the grating The half-width of the absorption / emission line becomes Here Δk g is the spectral resolution of the grating, k 0 is the central wave number. Therefore, the wave number width of the extracted absorption / emission line Δk ≥ 2×(ceil(3σ / k int )+1)×k int , where k int is the detection wavenumber range of a single pixel, where ceil(3σ / k int ) represents the conversion of 3σ / k int Round to the nearest integer greater than or equal to this value (3σ / k int is an integer).

[0037] Step 2: Use the low-pass filter characteristics of B-spline fitting to 3σ Perform B-spline fitting and normalize to get B nor .

[0038] Step 3: Use the improved EMD method to obtain the interference envelope BL. The specific steps of the improved EMD method are:

[0039] (I) Find B nor The maximum and minimum points of :

[0040] For interference information with very small contrast, the extreme value information is obtained by differentiation. Interference fringes are high-frequency components relative to the interference envelope, and high-frequency components can be enhanced by differentiation. The present invention uses B-spline fitting to achieve the approximation of numerical differentiation. First, B-spline fitting is performed on discrete data, and the fitting process can filter out abnormal points such as interference burrs and intermittent noise in the original signal. Then the second-order differential is calculated for the fitting result, and the minimum point of the second-order differential is used to represent the maximum point of the original interference signal, and the maximum point of the second-order differential represents the minimum point of the interference signal.

[0041] (ii) Determine whether the endpoint signal is an extreme value based on the periodic characteristics of the interference fringes, and then proceed to step (iii):

[0042] If it is an extreme value, determine whether it is a maximum value or a minimum value, and go directly to step (3);

[0043] If it is not an extreme value, in order to reduce the error caused by the endpoint effect, it is necessary to perform endpoint extension, and then determine whether it is a maximum value or a minimum value, and enter step (three). For the determination of the extension range, it is necessary to calculate the spacing between each extreme value. Due to the modulation of the interference fringes by the envelope, the distance between adjacent extreme values ​​is not equal. However, the interference fringes have a certain regularity with the change of the optical path difference, and for general detection light sources, their intensity has a certain symmetry. Here, it is necessary to design the range that needs to be extended according to the spacing characteristics between the extreme values. First, the ratio of the spacing between the adjacent maximum (minimum) and minimum (large) values ​​on the left and right sides is calculated respectively, and the ratio between each spacing is calculated. According to the symmetry, the corresponding ratios on the left and right sides of the interference signal are approximately equal. Therefore, the spacing between the last extreme value of the original signal and the extension extreme value point can be obtained according to the ratio and the previous spacing. It should be noted that since the extreme value point of the envelope does not completely coincide with the extreme value of the interference fringes, the obtained spacing needs to be corrected according to the difference between the corresponding spacing ratios known on the left and right sides. Then determine the extension range according to the distance between the original signal endpoint to the last (first) extreme value point and the spacing between the extreme value points obtained by the extension after correction.

[0044] The determination of extreme points can be divided into the following three cases:

[0045] When there are inflection points at both ends of the original signal, the extended extreme value can be solved by using the function value at the inflection point according to the characteristics of the interference fringes. Ideally, there must be an inflection point between adjacent maximum and minimum values, and the function value at the inflection point is the average of the maximum and minimum values. Therefore, the extended extreme value can be calculated using the function value at the inflection point and the adjacent extreme value.

[0046] When the side closest to the endpoint is an inflection point and the other side is an extreme value, the extreme value that needs to be extended on the side with the inflection point can be obtained according to the above method. For the situation on one side of the extreme value point, according to the symmetry, the ratio of the maximum (minimum) value to the minimum (large) value in reverse order can be calculated first, and then based on the obtained ratio, the piecewise cubic Hermite interpolation polynomial is used for extrapolation interpolation to obtain the next ratio, that is, the ratio of the extended maximum (minimum) value obtained on the side closest to the inflection point to the minimum (large) value that needs to be extended on the side closest to the endpoint. Then, based on this ratio and the extreme value obtained by extending the side of the inflection point, the extreme value that needs to be extended on the side of the extreme value can be obtained.

[0047] In the case where both sides closest to the endpoint are extreme values, firstly, the extreme value on either side is obtained by a method similar to the above-mentioned method where the extreme value is closest to one side of the endpoint, and then the ratio between the extreme values ​​symmetrical on the left and right sides of the original signal is calculated. Similarly, based on the relationship between the ratios, the piecewise cubic Hermite interpolation polynomial is used to perform extrapolation interpolation and extension to obtain the ratio between the extended extreme values, and finally, the extreme value to be extended on the other side is calculated based on the ratio and the extreme value on one side. Finally, the extreme value obtained by extension is merged with the extreme value of the original signal, and then it is determined whether it is a maximum value or a minimum value, and the process goes to step (iii).

[0048] In order to illustrate the influence of the envelope on the interference fringes, the present invention obtains the following equations based on the mathematical model of the coherent dispersion system shown in formula (1) and taking sampling into account: Figure 2 The absorption and emission lines shown correspond to B nor .

[0049]

[0050]

[0051] here is the light source intensity distribution function, Δτ is the half-maximum full width, τ is the optical path difference, k is the wave number, k n is the central wave number of the absorption / emission line, S 1 and S 2 are the characteristic parameters that reflect the intensity of absorption and emission lines. For absorption lines, S 1 =1, S 2 A negative number is the absorption depth. For emission lines, S 1 =0, S 2 A positive number is the emission intensity. Δk n and Δk g are the full width at half maximum of the absorption / emission line and the resolution of the grating, respectively.

[0052] Figure 2 is Δτ=3×10 -4 cm,τ=1.266cm,k n =1.5803×10 4 cm -1 , S 2 =0.75,Δk n =0.4995cm -1 , Δk g =0.7292cm -1 , the interference period is 5.2, and the dispersion direction sampling k int=0.015nm, the extreme value obtained after cubic spline difference fitting of the emission and absorption line interference information obtained by the 3σ rule. Where (a) and (b) represent the absorption and emission lines respectively, where the solid line is the interference information obtained by fitting, and "·", "*" and "+" represent the maximum, minimum and inflection point. It can be clearly seen from the figure that the distances between adjacent extreme points are different, and whether it is the emission line or the absorption line, the spacing from the maximum to the minimum on the left side of the curve showing an upward trend is smaller than the spacing from the minimum to the maximum, and the spacing from the maximum to the minimum on the right half is larger than the spacing from the minimum to the maximum, so the spacing of the extreme values ​​is approximately symmetrical, and the extension range can be determined based on this feature. Regarding the determination of extreme value points, here we give the case where the signal closest to the left endpoint is the extreme value, and the signal closest to the right endpoint is the inflection point. Therefore, the inflection point and the last extreme value information of the signal can be used to calculate the extreme value information required for the right extension. Then, based on the ratio of the maximum (minimum) value to the minimum (large) value in reverse order, and based on this ratio, the piecewise cubic Hermite interpolation polynomial is extrapolated and interpolated to obtain the ratio of the extreme value of the right extension to the extreme value of the left extension. The extreme value information of the right extension has been obtained based on the inflection point information, so the extreme value on the left can be obtained.

[0053] (iii) fitting the maximum (upper) and minimum (lower) envelope curves by cubic spline interpolation, and solving the average value M of the extreme envelope curves;

[0054] (IV) Using the original function B nor Subtract the mean value M of the envelope to get B′,

[0055] (V) Use Rilling stop condition to judge:

[0056] When the stop condition is met, the output envelope BL is output, BL = B nor -B′;

[0057] If the stopping condition is not met, let B nor =B′, continue to repeat the above steps (i) to (v) until the stop condition is met.

[0058] See also Figure 3 The method of the present invention is used to Figure 2 The envelopes extracted from the two images (a) and (b) are the envelopes obtained using the absorption line and the emission line, respectively. According to the parameters mentioned above, the interference images with line-of-sight velocities of 0 m / s and 1000 m / s are simulated, and the Figure 1 The process shown in the figure performs envelope correction on the interference signal, and the obtained result is as follows Figure 4As shown in the figure. Where "·" and "*" are the data of the line-of-sight velocity of 0m / s and 1000m / s respectively after correction, and the realization and dotted line resolution are the results of cubic spline difference fitting of the corresponding discrete data. It can be seen from the figure that after removing the envelope, the original interference information changes approximately in a sine and cosine manner with the optical path difference, and the fitting results show that the line-of-sight velocity accuracy obtained using the absorption line and the emission line is 1000.32m / s and 1000.24m / s respectively, and the RMSE is 0.56m / s and 0.14m / s respectively. Figure 5 It directly uses the data of the detector column (row) where the central wavelength of the absorption / emission line is located, and uses the extreme value information to invert the radial velocity result. Figure 5 and Figure 4 From the two figures, it can be seen that the accuracy of the radial velocity obtained by inversion is greatly improved after adopting the 3σ rule and removing the envelope.

[0059] The above is an explanation of the specific implementation of the present invention, rather than a limitation of the present invention; technicians in the relevant technical field can also make various equivalent technical solutions without departing from the scope of the present invention, so all equivalent technical solutions should be included in the protection scope of the present invention.

Claims

1. An improved EMD interference envelope extraction method based on extrapolation interpolation, characterized in that: The following steps are involved: Step 1: Input the interference image B to be processed and extract the interference information B of the absorption / emission line according to the 3σ rule 3σ ; Step 2: Use the low-pass filter characteristics of B-spline fitting to 3σ Perform B-spline fitting and normalize to get B nor； Step 3: Obtain the interference envelope BL using the improved EMD method. The specific steps of the improved EMD method are: (I) Use the differential method to find the B that is submerged due to envelope modulation nor The maximum and minimum points of ; (ii) judging whether the endpoint signal is an extreme value according to the periodic characteristics of the interference fringes, and if it is an extreme value, judging whether it is a maximum value or a minimum value, and proceeding to step (iii); if it is not an extreme value, after performing endpoint extension, judging whether it is a maximum value or a minimum value, and proceeding to step (iii); (iii) fitting the maximum (upper) and minimum (lower) envelope curves by cubic spline interpolation, and solving the average value M of the extreme envelope curves; (IV) Using the original function B nor Subtract the mean value M of the envelope to get B′; (V) Use Rilling stop condition to judge: When the stop condition is met, the output envelope BL is output, BL = B nor -B′; If the stopping condition is not met, let B nor =B′, continue to repeat the above steps (i) to (v) until the stop condition is met.

2. The improved EMD interference envelope extraction method based on extrapolation and interpolation according to claim 1 is characterized in that: In step 1, the wave number width of the extracted absorption / emission line Δk≥2×(ceil(3σ / k int )+1)×k int , where k int is the detection wavenumber range of a single pixel, where ceil(3σ / k int ) represents the conversion of 3σ / k int Rounds to the nearest integer greater than or equal to the value.

3. The improved EMD interference envelope extraction method based on extrapolation and interpolation according to claim 2 is characterized in that: In the step three (a), firstly, B-spline fitting is performed on the discrete data, and the fitting process can filter out abnormal points such as interference burrs and intermittent noise in the original signal; then the second-order differential is calculated for the fitting result, and the minimum point of the second-order differential is used to represent the maximum point of the original interference signal, and the maximum point of the second-order differential represents the minimum point of the interference signal.

4. The improved EMD interference envelope extraction method based on extrapolation and interpolation according to claim 3 is characterized in that: In step 3 (ii), the specific steps of performing endpoint extension are: firstly calculating the spacing between each extreme value, then determining the spacing between the extreme value points of the original signal and the extended extreme value points by using the ratio relationship of the extreme value spacing according to the periodicity of the interference fringes and the symmetry of the light intensity envelope, and finally determining the extension range according to the distance between the last (first) extreme value point of the original signal and the endpoint; For the extended extreme values, when there is an inflection point near the endpoint, the function value at the inflection point is used for calculation according to the periodicity of the interference fringes and the symmetry of the light intensity envelope; when there is an extreme value near the endpoint, the ratio of the maximum (minimum) value to the minimum (large) value is calculated first, and then the ratio is extrapolated and interpolated, and finally the extended extreme value information is determined by combining the extrapolation result and the extreme values ​​at both ends of the signal.

Images