36results about How to "Improve decomposition accuracy" patented technology

The invention provides non-invasive load monitoring method and system based on current decomposition. The method includes the steps of firstly, training each electrical device to obtain statistic manifolds of the electrical device respectively, and storing the statistic manifolds into a database; secondly, installing an acquisition unit at an inlet of a power grid, acquiring electrical data of all the electrical devices in real time, and converting the electrical data into vector representation of calculating statistic manifolds to acquire a vector set of total current; and thirdly, calculating statistic manifolds of the unit loading in the electrical devices respectively, and decomposing a total statistic manifold according to the statistic manifolds to acquire real-time current vectors of the electrical devices respectively. Acquired total current is directly decomposed to each device, detailed electrical information on related devices is calculated by current information directly instead of evaluating complex operation status, and accordingly accurate and detailed electrical information can be acquired.

The invention discloses a vegetation coverage estimation method based on an improved linear spectrum mixed model. The method includes: obtaining image data in the research area; performing rough geometric correction and radiation correction preprocessing on the obtained image data; using the pure pixel index method to extract endmembers from the preprocessed images, and constructing variable endmembers Linear spectral decomposition model; according to the constructed variable endmember linear spectral decomposition model, the vegetation coverage information in the research area is extracted. By applying the present invention, the estimation accuracy of vegetation coverage can be improved.

The invention relates to a method for iteratively extracting the maximum energy of a marine structure measured signal, comprising: calculating the autocorrelation function of the original dynamic response signal, solving the power spectral density, and determining the maximum energy component in the original dynamic response signal; discretizing the original dynamic response signal , and perform complex exponential fitting under the low-order state-space model to extract the maximum energy component in the original dynamic response signal; set the decomposition stop criterion, increase the order of the state-space model, and repeat the decomposition process to quantify the decomposition accuracy; The maximum component component is constructed, and the remaining dynamic response signal is used to replace the original dynamic response signal, the above decomposition process is repeated, and the maximum energy component of the remaining dynamic response signal is iteratively extracted to achieve a complete decomposition of the original dynamic response signal. The maximum energy component in each decomposition is determined based on the power spectral density function of the dynamic response, and the maximum energy component in the response is extracted iteratively by controlling the extraction precision, which improves the decomposition precision.

The invention discloses a GNSS multipath effect suppression method based on EMD iteration threshold value smoothing. EMD is performed on measuring errors to obtain N intrinsic mode function components, multipath errors are often concentrated in a high-order IMF and a remainder item, on one hand, singular spectrum analysis is used for performing noise estimation on N IMFs after EMD, and according to estimation results, first M IMFs are selected for noise auxiliary data analysis, so that EMD results of the measuring errors are obtained; on the other hand, the mode that first IMF data are sampled randomly is adopted, multiple sequences which are the same as an original signal in signal-to-noise ratio are constructed, mode unit threshold value smoothing is adopted for removing noise of the sequences, multiple results after noise removal are averaged to obtain the GNSS multipath errors through separation, and the GNSS multipath errors are used for compensating GNSS measuring results to achieve multipath effect suppression in the measuring process. According to the GNSS multipath effect suppression method, EMD precision is improved, and influences of noise are reduced.

This disclosure proposes a channel wave signal multipath spectrum analysis method and system. Multiple channel wave signals are obtained by transmitting detection signals at different angles, which can achieve the purpose of multi-directional detection of the target geological interface, increase the detection range, and more To reflect the geological information of the formation to be measured, in terms of channel wave signal extraction, a non-orthogonal decomposition method is used to decompose the channel wave signal to decompose each multipath in the channel wave signal, and multiple echo signals are fused to obtain the channel wave signal. The multipath spectrum of the wave signal. This decomposition method has high decomposition precision and is more practical. The decomposed signal can retain as many characteristics of the interface signal as possible, and can express the obtained signal more concisely, thereby improving the signal-to-noise ratio of the signal, reducing the bit error rate of the signal, and improving the accuracy of slot wave detection.

The invention discloses an ensemble empirical mode decomposition (EEMD) method having an excitation noise add parameter option function, relates to the technical field of signal analysis and signal processing, and solves the problem that the conventional EEMD method cannot implement high decomposition accuracy and less calculated amount simultaneously when nonlinear and non-stationary signals areprocessed. The method comprises the following steps of: 1, setting initial values of an ensemble number and a noise amplitude; 2, performing the EEMD on a signal to acquire an internal stability modefunction matrix; 3, solving and comparing the lower bound of a resolution error with the previous result to judge whether the resolution error is reduced so as to determine whether the noise amplitude needs to be further reduced; 4, changing the ensemble number to acquire a resolution error of a new ensemble number; and 5, comparing the resolution error to ensure that the difference of the resolution errors of the EEMD is less than that of the preset resolution errors, and stopping the decomposition to finish the EEMD having an excitation noise add ensemble number and an add amplitude option function. The ensemble empirical mode decomposition method is suitable for the processing of the nonlinear and non-stationary signals.

The invention provides an automatic decomposition method of an array type sEMG (surface EMG) signal. The automatic decomposition method comprises the following steps: firstly, pretreating the array type sEMG signal, and calculating an initial distribution sequence vector; determining the number of kinematic units, namely, the cluster number, by using time domain subtraction; according to the cluster number, clustering array type sEMG waveforms by using a minimum distance classifier; finally, recalculating a new kinematic unit distribution sequence vector, circulating the executive routine till decomposition is finished, classifying all the distribution sequences, and optimizing the result. The decomposition method is high in accuracy, quick in calculation speed and easy to operate.

The invention provides a method for identification and separation of magnetotelluric signal and noise. Firstly, the magnetotelluric signal collected in the environment of strong electromagnetic interference is segmented at equal intervals, and the optimal intrinsic time scale decomposition is used to adaptively decompose each segment of the magnetotelluric signal to obtain the rotation component, and the sample entropy and fuzzy entropy of the rotation component are extracted . Then, the sample entropy and fuzzy entropy are used as the joint characteristic parameters for fuzzy C-means clustering. According to the characteristic parameters and clustering method, the measured magnetotelluric sequences are divided into two categories: useful signals and strong interference signals. Finally, only for the magnetotelluric signal segments identified as strong interference signals, the wavelet threshold is used for noise suppression processing, and the reconstructed magnetotelluric signal is obtained by superposition. The invention has high reliability.