127 results about "Singular spectrum analysis" patented technology

The invention discloses a heart rate estimation method and device for wearable heart rate monitoring equipment. The heart rate estimation method mainly includes: removing motion artifact and tracking heart rate spectrum peak, where motion artifact removing includes: utilizing a nonlinear self-adaptive filter method to capture a nonlinear relation between noise reference signals and motion artifact noise in pulse wave signals so as to effectively eliminate motion artifact interference, adopting a binary decision-making method based on classification to judge whether filtered pulse wave signals still contain a lot of noise or not, and adopting a singular spectrum analysis method to further eliminate noise interference of the pulse wave signals still containing noise; heart rate spectrum peak tracking based on frequency spectra includes: positioning heart rate spectrum peak of each time window, namely positioning the heart rate spectrum peak on the basis of a nonlinear positioning method, and positioning the heart rate spectrum peak on the basis of a classification positioning method if the nonlinear positioning method fails. The heart rate estimation method is used for heart rate estimation and is high in calculating accuracy and low in complexity, so that enforceability of the wearable monitoring equipment is guaranteed.

A network traffic shaper adjusts the rate of data flowing from a packet source to a packet sink responsive to priority assigned to the packets pursuant to their ability to match a series of filters. Each packet is applied to a queue depending on its classification. The packets are read from each queue in accordance with a queue reading policy. Singular spectrum analysis of the traffic load produces a set of vectors that capture the spectral bases of the traffic state. These vectors are associated with the bandwidth demanded by the traffic, a relationship that is recorded in a queue / time / eigenvector / bandwidth / control table. Samples of the current network load are compared with previously recorded vector bases to reveal how similar the current traffic state is to previous traffic states, which in turn predicts the traffic's bandwidth needs for the immediate future. The queue reading regime is updated in response to this prediction.

The invention relates to an establishing method of a regional CORS coordinate time series noise model. The establishing method is characterized by comprising the following steps that: step 1, observation data of a regional CORS station are obtained; step 2, GNSS data pre-processing software TEQC is adopted; step 3, GAMIT / GLOBK baselines are adopted to solve adjustment; step 4, a high-precision coordinate time series is established; step 5, singular spectrum analysis-based spatial filtering is performed; step 6, MLE estimation-based noise model selection is performed; and step 7, regional CORS velocity filed estimation is performed in view of colored noise models. According to the establishing method of the invention, time-domain frequency domain characteristic analysis is adopted to perform spatial filtering on the regional CORS time series; and maximum likelihood values are selected for different noise models of the regional CORS time series, so that judgment can be performed, and therefore, optimal noise models in the N (North), E (East) and U (altitude) are determined, and the regional CORS velocity filed can be estimated in view of the optimal noise models.

The invention discloses a wrist heart rate monitoring method of a sports bracelet. The method comprises the steps that firstly, threshold processing is conducted on an original optical pulse wave signal, and signals not relevant to the heart rate are eliminated; then, a synchronous acceleration signal is processed through a singular spectrum analysis method, and an motion artifact reference signal is obtained; the motion artifact signal and a preprocessing optical signal are subjected to self-adaption filtering processing, motion artifact components in the optical signal are eliminated, and a pure pulse wave signal is obtained; the pure pulse wave signal is processed through a self-adaption threshold iteration method, and the frequency spectrum precision is improved; finally, sufficiency analysis is conducted on situations possibly occurring in peak value selection of a pulse wave frequency spectrum, and then the pulse wave peak value is determined. By accurately determining the peak value and elapsed time, the heart rate value of current time can be calculated. By means of the method, interference of motion artifact to the optical pulse wave signal can be effectively reduced, and then the precision of heart rate estimation is improved.

The invention provides an ultrasonic echo signal online preprocessing method based on singular spectrum analysis. The method is used for echo signal preprocessing in the ultrasonic online detection technology. The method includes the steps that a rail matrix is constructed according to measured ultrasonic echo signals, and proper matrix parameters are automatically determined through the inset window length selection algorithm; the ultrasonic echo signal rail matrix is subjected to singular value decomposition, and a singular value spectrum is obtained; a feature value spectrum and a singular value differential spectrum of the ultrasonic echo signal rail matrix are calculated; a reconstruction group number for singular spectrum analysis is determined through a self-adaptive reconstruction algorithm, and then a denoised reconstruction ultrasonic echo sequence is constructed; a reconstruction principal component sequence corresponding to a signal component is further subjected to frequency domain analysis.

The invention discloses a GNSS base station crustal movement velocity estimation method in consideration of nonlinear change. In consideration of the influence on the base station velocity estimationby the GNSS base station nonlinear change, a random model and like, the latest resolving strategy is adopted to acquire a GNSS base station coordinate time sequence under an ITRF2014 framework, a timesequence model for performing separation on the nonlinear change of the base station coordinate time sequence is established by combining a real physical correction model, abnormal location gross error and step detection and like methods; and a GNSS base station coordinate time sequence background noise model determination technology based on the singular spectrum analysis method is adopted, andan improved Bayes information minimum criterion noise model estimation method is proposed base on above conditions, a GNSS base station crustal movement velocity field estimation method is provided, thereby providing accurate and reliable velocity field data basis for the crustal movement velocity field estimation.

The invention relates to a frequency chosen of singular-spectrum analysis-based magnetic resonance sounding signal extraction method. According to the method, a nuclear magnetic resonance sounding water detector is used to collect MRS signals of a Larmor frequency-known region; a broadband band-pass filter is adopted to suppress partial noises, and thereafter, the position of MRS signals corresponding to the Larmor frequency are found on a power spectrum based on power spectrum analysis; and frequency chosen of singular-spectrum analysis is performed, so that the MRS signals are extracted. Thefrequency chosen of singular-spectrum analysis includes the following four steps that: embedding is performed; RSVD (Regularized Singular Value Decomposition) is performed; and corresponding singularvalues are selected according to the amplitudes of the MRS signals so as to perform matrix reconstruction; and diagonal averaging is performed. With the method of the invention adopted, random noises, spike noises and power frequency harmonic interference can be effectively filtered from noise-containing MRS signals; and the effective extraction of the MRS signals under a complex strong noise interference condition can be realized. Compared with a traditional MRS signal de-noising method, the method has the advantages of high operation speed, high signal-to-noise ratio, high practicality andthe like.

The invention discloses an ionosphere TEC (Total Electron Content) anomaly detection method, which mainly comprises the steps of forming an ionosphere TEC time sequence, performing decomposition and reconstruction by using SSA calculate a background value, adopting a sliding quartile range method for an absolute value of the difference of the background value and an observed value to calculate a tolerance value at each moment of the detection day, and thus calculating an upper limit value and a lower limit value of the detection day. Singular spectrum analysis is adopted and slides day by day to calculate ionosphere TEC anomalies on the other days. The ionosphere TEC anomaly detection method performs ionosphere IEC anomaly detection based on a sliding singular spectrum analysis method, not only considers the background value of the ionosphere TEC at the detection moment, but also applies the traditional sliding quartile range method to the absolute value of the difference of the time sequence observed value and the background value to calculate the upper limit value and the lower limit value at each moment of the detection day, so that the accuracy and precision of an ionosphere TEC anomaly detection result are greatly improved by using robust statistical mathematical features of the sliding quartile range method, and the ionosphere TEC anomaly detection method has a characteristic of being universal.

The invention discloses a data preprocessing method of a water quality prediction system. The method comprises the following steps: 1) each monitoring station loads water quality data to a monitoring database, a user sends a user instruction to a network interface, a network server converts the user instruction into a structured query language (SQL) instruction and sends the SQL instruction to the monitoring database and a water quality database, and the data of the monitoring database is returned to the network server; 2) the water quality data is loaded into the memory of the network server in the form of a table in the monitoring database and the water quality database, and the data structure of the data to be analyzed is established; and 3) k-means clustering analysis is firstly performed on the water quality data, then the missing data is interpolated, and finally singular spectrum analysis is performed, thereby reducing the data noise and extracting an oscillation component and a trend component. The method disclosed by the invention reads the related water quality data from the database more quickly, improves the prediction efficiency, provides clean, accurate and concise data to a prediction model, and improves the prediction accuracy and stability.

The invention belongs to the technical field of information retrieval and database structures, and discloses a surface roughness monitoring model based on data mining and a construction method. The surface roughness model is established based on variance analysis and regression analysis, the incidence relation between cutting force and vibration signals and the surface roughness is determined according to the clustering result, and blindness in the cutting signal selection process is greatly reduced. A multi-sensor technology is applied, force and vibration signals in the cutting process are collected in real time, the cutting signals are decomposed and reconstructed on the basis of singular spectrum analysis, interference generated by noise signals can be effectively reduced, and characteristic quantity extraction is facilitated. Time domain and frequency domain results of cutting force and vibration signals are analyzed, feature extraction is carried out through correlation selection, a surface roughness prediction model is established through a radial basis function neural network, prediction precision and the intelligent level can be greatly improved, and online real-time prediction can be achieved.

The invention discloses an intelligent rotary machine fault depth network feature identification method. A vibration sensor is arranged at a to-be-detected rotating mechanical part of a train rollingbearing; collecting an original vibration sequence when the rolling bearing works; decomposing and reconstructing the original vibration sequence through a singular spectrum analysis method; extracting a root-mean-square value of the reconstructed vibration sequence; standard deviation, skewness index and peak value; a fault position is judged by using a rotary machine fault position diagnosis model obtained by training of a support vector machine; and then, carrying out ensemble empirical mode decomposition on the reconstructed vibration sequence, calculating the permutation entropy values ofa group of decomposed intrinsic mode components, taking the permutation combination of the permutation entropy values as a detection characteristic, and judging the fault type by using a rotary machine fault type diagnosis model obtained by training of a support vector machine. The fault position and the fault type of the rotary machine can be detected more timely, and the fault diagnosis accuracy and reliability are improved.

The invention relates to a short-term wind power forecasting method based on a hybrid algorithm, which includes the following steps: S1 decomposing the original wind power into a series of intrinsic mode function (IMF) sub-modal components by using the integrated empirical mode decomposition technique, S2 extracting the main trend components of IMF and RES components except the first IMF componentIMF1 decomposed by the integrated empirical mode decomposition (IMD) technique by using the singular spectrum analysis (SSA) method to obtain more obvious sub-modal components, S3 preserving IMF1 andR, and decomposing IMF1 and R into a series of stationary sub-modal components by wavelet packet decomposition. S4 making use of on-line robust limit learning machine to to establish a prediction model for all the sub-modes obtained in the step S1-S3,, and obtaining a final wind power prediction result by superposing the sub-modes; The invention can effectively and accurately predict the actual wind power system, and provides an important reference for the operation and planning of the electric power system.

The invention discloses a noninvasive blood sugar data processing method and a noninvasive blood sugar data processing system based on a convolutional neural network. The method comprises the steps of acquiring a plurality of sets of blood sugar data; performing calculation for acquiring a maximum infrared signal; through singular spectrum analysis and empirical mode decomposition, performing decomposition, grouping and ordering on the maximum infrared signal; respectively extracting the maximum infrared signal, the average value, the variance, the slope and the peak value of front N sets of component data, thereby constructing a characteristic signal; according to the characteristic signal and the blood sugar value of a plurality of sets of blood sugar data, constructing a mapping matrix; according to a to-be-tested signal of a to-be-tested person and the mapping matrix, constructing a to-be-tested mapping matrix; by means of a characteristic mapping layer and a pooling layer of the convolutional neural network, optimizing the to-be-tested mapping matrix, and outputting an optimization result, wherein a radial primary function is used as an activating function in a characteristic mapping layer; and the pooling layer is used for reducing the number of dimensions of the signal. The noninvasive blood sugar data processing method and the noninvasive blood sugar data processing system can improve blood sugar data estimation precision.

The invention relates to the field of short-term wind speed prediction, and discloses a short-term wind speed prediction method based on SSA-HMD-CNNSVM model. The method includes: firstly, utilizing singular spectrum analysis (SSA) to reduce noise and extract trend information of original wind speed data; and then, carrying out deep decomposition on the wind speed data by using mixed mode decomposition, then, predicting each wind speed sub-layer by using a convolutional support vector machine, and finally, carrying out superposition on prediction results of all components, thereby obtaining afinal wind speed prediction result. Compared with a common signal preprocessing mode, noise reduction and deep decomposition processing on the wind speed time sequence can effectively reduce the influence of random fluctuation of the wind speed time sequence on a prediction result, and the accuracy and precision of wind speed prediction are greatly improved. Meanwhile, the CNNSVM can combine the advantages of a single model convolutional neural network and a support vector machine, so that the wind speed prediction method has strong generalization capability and robustness, and can be appliedto wind power plant wind speed prediction on a large scale.

The invention discloses a wind power prediction method based on singular spectrum analysis and deep learning. The method comprises the following steps: obtaining wind power, wind speed and wind direction historical data, and preprocessing the wind power, wind speed and wind direction historical data to obtain a wind power, wind speed and wind direction angle time sequence; taking a sine value anda cosine value of the wind direction angle time sequence; utilizing singular spectrum analysis to extract trend components and oscillation components of the wind power and wind speed time series, andreconstructing the two components; splicing the reconstructed sequence with the sine of the wind direction and the cosine of the wind direction to form an m@T * n tensor; dynamically selecting a training sample, and establishing a convolutional neural network-gated cycle unit deep learning prediction model; and predicting the generated tensor by adopting a convolutional neural network-gated cycleunit deep learning prediction model to obtain a predicted wind power time sequence. According to the method, the reconstruction time sequence of noise reduction is obtained through singular spectrum analysis, and the prediction precision is further improved.

The invention discloses a short-term photovoltaic decomposition prediction method considering meteorological factor changes. The short-term photovoltaic decomposition prediction method comprises the steps that S1 a photovoltaic output time sequence is decomposed through a singular spectrum analysis method so as to obtain a low frequency sequence, a high frequency sequence and a noise sequence; S2main meteorological factors influencing the photovoltaic output are determined by using the Pearson correlation coefficient method and the sensitivity of the main meteorological factors for the photovoltaic output is analyzed; S3 a prediction model considering the meteorological factors is established by aiming at the low frequency sequence and the high frequency sequence with combination of the sensitivity; and S4 the low frequency sequence prediction value and the high frequency sequence prediction value are acquired according to the prediction mode, and the photovoltaic output prediction value is acquired according to the low frequency sequence prediction value and the high frequency sequence prediction value. The photovoltaic output is decomposed into different subsequences through thesingular spectrum analysis method to independently analyze the characteristics of each sequence; and the influence degree of unit change of different meteorological factors on the photovoltaic outputis acquired through correlation analysis and sensitivity analysis so as to more accurately predict the photovoltaic output.

The invention relates to a seismic irregular noise removing method based on robust singular spectrum analysis. The seismic irregular noise removing method is characterized by comprising the followingsteps that the single frequency seismic data are embedded in the Hankel matrix so as to construct the Hankel matrix of the single frequency seismic data; the objective function of the Hankel matrix ofthe single frequency seismic data is obtained based on the L1 and L2 mixed norm; the objective function is solved according to the preset regularization parameters in the objective function so as toobtain singular spectrum analysis of the Hankel matrix based on the L1 and L2 mixed norm; and alternating minimization and weighted least square minimization are performed on singular spectrum analysis in turn to solve the factor matrix of the Hankel matrix of the single frequency seismic data so as to obtain estimation of the Hankel matrix of the single frequency seismic data and complete noise reduction of the single frequency seismic data. The seismic irregular noise removing method based on robust singular spectrum analysis can be widely applied to the field of seismic data processing.

The invention discloses an urban air quality time sequence prediction method considering space-time relevance. According to the method, singular spectrum analysis is introduced to carry out time sequence prediction on PM2.5 monitoring data and meteorological characteristic data; a space-time correlation cube is designed to adaptively select first K important spatial neighborhood site features, a time sequence prediction result and the first K important spatial neighborhood site features are superposed to construct a sample feature set, and finally fitting of final results under different timescales is completed by using a random forest algorithm. Through the coupling model provided by the invention, the space-time relevance between different space stations can be effectively considered, so that the time sequence prediction effect and the stability degree of a single station in an urban space environment under different time scales are improved, and a reference basis can be provided for urban atmosphere management decision making.

The invention discloses a short-term impact load prediction method based on a two-layer decomposition technology. The short-term impact load prediction method comprises the following steps of obtaining impact load historical data and performing equalization preprocessing on the data; decomposing the preprocessed impact load historical data into a plurality of discrete modal components through a variable mode, and recording the discrete modal components as IMFn, where n is a serial number of the discrete modal components; performing secondary decomposition on the component with the highest frequency in the discrete modal components by singular spectrum analysis to obtain a plurality of sub-sequences; constructing an extreme learning machine neural network prediction model based on whale algorithm optimization; inputting the components except the component with the highest frequency in the discrete modal components and a sub-sequence obtained by secondary decomposition into an extreme learning machine neural network prediction model based on whale algorithm optimization; and superposing prediction values output by the extreme learning machine neural network prediction model based onwhale algorithm optimization to obtain an actual prediction result. According to the method, the influence of nonlinear characteristics in the impact load is overcome, and the prediction precision iseffectively improved.

Systems and methods for analysis of data associated with fluid flow through a conduit and potential control of the fluid flow. In one embodiment, a method includes a data processing unit obtaining from a sensor a signal representing a series of measurements of a physical parameter associated with fluid flow through the conduit. The data processing unit performs singular spectral analysis of this signal and thereby identifies recurring, possibly non-sinusoidal patterns in the signal. The data processing unit may construct a second signal from some or all of the recurring patterns identified in the first signal. Alternatively, the data processing unit may identify recurring patterns in the signal which are indicative of corresponding conditions associated with the fluid flow through the conduit. The data processing unit may then initiate control functions responsive to the recurring patterns and associated conditions.

The invention discloses a gas load combination prediction method based on support vector regression and relates to gas load prediction methods. According to the combination prediction method, a data preprocessing technology, an improved genetic algorithm and support vector regression are combined, and the method is mainly used for solving the problems that in the prior art, urban gas load prediction is low in precision and poor in applicability. The method comprises the steps that first, a correlation coefficient method is adopted to analyze the correlation between different influence factorsand gas loads, and singular spectrum analysis is adopted to perform de-noising processing on the obtained main influence factors; second, processed data is adopted to train a support vector regressionmodel, nuclear parameters and penalty factors are optimized in combination with the improved genetic algorithm, and finally a support vector regression model with an optimal training result is obtained; and last, the trained support vector regression model is utilized to predict gas load indexes in a future period of time. Through the combination prediction method, a short-term gas load prediction error can be substantially lowered, and prediction precision can be improved.

The invention discloses an electricity price super short-term prediction method. The method comprises the steps that S1, electricity price historical data is acquired, and the electricity price historical data is preprocessed to obtain an electricity price sequence; S2, singular spectrum analysis is utilized to directly extract trend components, oscillation components and noise components in the electricity price sequence; S3, the noise components are filtered out, and a singular spectrum sequence is adopted to perform reconstruction on a trend component and oscillation component sequence obtained after noise reduction to obtain training samples; S4, the training samples are dynamically selected, and a prediction model of a cuckoo search algorithm optimized extreme learning machine is established; S5, the cuckoo search algorithm optimized extreme learning machine model is adopted to perform 0.5h-advanced prediction on the trend component and the oscillation component sequence to obtainsub-sequences; and S6, prediction values of all the sub-sequences are added to obtain an actual prediction result. According to the method, singular spectrum analysis is adopted to preprocess the original data, an input weight and hidden layer offset of the cuckoo search algorithm optimized extreme learning machine are adopted, and a non-optimal parameter defect of the extreme learning machine iseffectively overcome.

The invention provides a pipe network leakage warning method based on a water consumption optimization prediction model. The pipe network leakage warning method comprises the following steps of obtaining original data; preprocessing historical water consumption sequences; analyzing mutual information between the preprocessed historical water consumption sequences; establishing a water consumptionprediction model when a support vector returns; and using a test set residual sequence to establish an isolated forest model, and judging the anomaly based on the real-time residual. According to thewater consumption optimization prediction model and the pipe network leakage warning method, the historical water consumption is processed based on L1 norm singular spectrum analysis interpolation, and the original sequence is restored as much as possible to improve the prediction accuracy; and the isolated forest algorithm is adopted for conducting abnormality identification, and fast and efficient online pipe network leakage warning can be achieved.

The invention discloses a singular spectrum analysis-based landslide mass displacement prediction method. The method is specifically implemented according to the following steps of performing data preprocessing on a time sequence by utilizing a spectral decomposition theory and an embedded reconstruction theory of singular spectrum analysis to obtain the accumulated landslide displacement data; removing the trend term displacement from the accumulated displacement to obtain the periodic term displacement; adopting Gaussian fitting to perform fitting prediction on the trend term displacement; selecting influence factors from the predicted trend term displacement by adopting a rapid multi-principal-component parallel extraction algorithm, and selecting the LSTM model related parameters by utilizing a Bayesian optimization algorithm; constructing a training set, a verification set and a prediction set, and establishing an LSTM network model to predict the periodic item displacement; and according to a time sequence decomposition principle, superposing the predicted values of the displacement sub-sequences to obtain a final predicted value of the displacement, thereby finishing the landslide body displacement prediction method. According to the present invention, the problem that multi-source heterogeneous influence factors are difficult to fuse for collaborative and dynamic forecasting in the prior art, is solved.

The invention discloses a method for simultaneously removing multiple artifacts from single-channel electroencephalogram signals. The method includes the following steps that, firstly, single-channelelectroencephalogram signals are decomposed through singular spectrum analysis, and a multi-dimensional signal component matrix is obtained; 2, a second-order blind identification method is used for blind source separation of the signal component matrix, and several independent source components are obtained; 3, the independent source components containing the artifacts are selected and subjectedto zero setting; 4, the independent source components obtained after zero setting are subjected to blind source separation inverse transformation, and the clean single-channel electroencephalogram signals are obtained after reconstruction. The method achieves simultaneous removal of the artifacts (myoelectricity, electro-oculogram and electrocardio artifacts) from the single-channel electroencephalogram signals, and electroencephalography components in the signals are effectively retained.

The invention discloses a short-term wind power prediction method based on double-time sequence feature learning, and the method comprises the following steps: building a training set and a test set,and converting original data into labeled data at the same time; Adopting a singular spectrum analysis method to perform de-noising and principal component selection on the original wind power data; Constructing a double-time-sequence feature learning neural network model composed of a local time sequence learning module and two long-short-term memory networks, and obtaining local wind power dataat different moments according to the input of the neural network model; And the neural network model outputs the double-time sequence characteristics processed by one local time sequence learning module and two long and short term memory networks through a full connection layer, and performs final regression analysis to obtain a to-be-predicted wind power value at the t + 1 moment at the t moment. According to the method, through principal component selection and multi-scale time sequence characteristic learning of original data, accurate prediction of the power generation power of the singlefan of the wind power plant is finally realized.

The invention provides an on-line prediction method of micro-milling chatter. A monitoring signal in a micro-milling machining process is obtained through a sensor, wherein the monitoring signal comprises a cutting force signal, a sound signal and an accelerated speed signal. The on-line prediction method of the micro-milling chatter comprehensively utilizes the milling force, the accelerated speed and the sound signal of the micro-milling machining process to represent a machining state, the on-line monitoring of the micro-milling machining process and the early prediction of the milling chatter are realized through a multichannel singular spectrum analysis algorithm and a GHMM (General Hidden Markov Model), a necessary technical means is provided for chatter inhibition research, so that a monitoring system module is greatly simplified, and the instantaneity of a machining state monitoring system is obviously improved.

The invention relates to a sea level change nonlinear trend extraction method, which comprises the following steps of decomposing a sea level change time sequence by utilizing an empirical mode decomposition method to obtain an intrinsic mode function and a trend term with different frequencies; eliminating a low-frequency false component by using a modal function method, and marking and eliminating a high-frequency noise component by using a frequency divergence method; sequentially carrying out frequency spectrum analysis by adopting a Fourier method to obtain frequencies corresponding to the peak values in each component frequency spectrogram, removing the invalid frequencies according to a Nyquist theorem, and converting the frequencies into periods; obtaining all possible embedded calculation windows, through the singular spectrum analysis, taking the trend term of empirical mode decomposition as a reference, and selecting the trend term with the minimum difference as the final nonlinear trend of the sea level change. According to the method, the automatic selection of the optimal window and the automatic extraction of the optimal sea level change trend can be realized, the adaptability is good, the efficiency is high, the trend extraction is stable, and the influence of the time sequence length is small.

The invention discloses a weak signal blind extraction method, and aims to provide a blind extraction method with more excellent robustness and anti-interference performance, and the method is realized through the following technical scheme: building a single-channel instantaneous linear hybrid model on the premise of independent signal sources; constructing the single-channel mixed signal into amatrix C of which the window length L is 2, performing singular value decomposition on the matrix C according to a singular spectrum analysis theory, and obtaining a time sequence corresponding to a singular value through inverse Hankel transform; continuously performing MRSSA transformation, and converting the single-channel instantaneous linear hybrid model into a multi-channel common-channel instantaneous linear hybrid model to obtain a reconstruction matrix X; after reconstruction, contructing an MRSSA-ICA filter, decomposing single-channel mixed data into a plurality of ICs, extracting interested signal components from single-channel time-frequency partially overlapped signals, and achieving common-channel blind extraction of weak signals based on MRSSA-ICA.

A singular spectrum analysis-based ionized layer anomaly detection method provided by the present invention comprises the steps of obtaining the historical ionized layer observation data ION; utilizing a singular spectrum analysis method to obtain an ionized layer normal variation component IONmain; calculating the normal background noise epsilon based on an ionized layer quiet period; according to the ionized layer normal variation component IONmain and the normal background noise epsilon, obtaining a normal variation range of the ionized layer observation data ION; when the actual ionized layer observation data exceeds the normal variation range of the ionized layer observation data ION, representing that an ionized layer is abnormal. According to the present invention, by the singular spectrum analysis, a periodic signal can be identified and reinforced, the extracted normal variation component contains the influences, such as the season change of the ionized layer along with the revolution of the earth, the 9-day, 13.5-day and 27-day periodic variation of the ionized layer caused by a sun 27-day rotation period, etc., so that the interferences caused by different external environments of the earth at a background field time period and the anomaly detection time period are removed from an anomaly detection result, and the ionized layer anomaly information can be detected effectively.