39 results about "Signal clustering" patented technology

Methods of generating an extended cluster are disclosed. A first cluster including data representative of a first signal indicative of an abnormal physiological symptom is generated. A second signal is detected as representing a second abnormal physiological symptom and the second abnormal physiological symptom continues after the first abnormal physiological symptom ends. An extended cluster including the data from the first signal and the second signal is generated that extends from the time when the first abnormal physiological symptom occurs to the time when the second abnormal physiological symptom ends. The first and the extended cluster may include data of both the first and second signals the entire period of the clusters. If desired, the first signal or the second signal may be provided from a sensor implanted in a patient's brain tissue.

In order to solve the technical problem that working performance of photo counting detector-adopted three-dimensional imaging laser radars is reduced under strong noise environments, the invention provides a photo counting-based three-dimensional image signal-noise separation and mixed regularization reconstruction method. The method comprises the following steps of: determining a signal cluster set on each pixel by adoption of a sliding time window method and filtering noise counting except time windows of the signal cluster sets; if both signal counting and noise counting do not exist on a certain pixel, constructing a superpixel by replacing a time accumulation process by space accumulation by means of counting information on a neighborhood pixel, and filtering the noise by adoption ofthe sliding tine window method; and finally, reprocessing a preliminary noise-filtered image by utilizing a total variation regular item, wavelet regular item and contourlet regular item combined mixed regularization method, so as to finally obtain a high-quality three-dimensional image. According to the method, signal photos can be effectively separated from noises under strong background noises,so that the rapid and accurate reconstruction of target three-dimensional images is realized.

A generalized pattern recognition is used to identify faults in machine condition monitoring. Pattern clusters are identified in operating data. A classifier is trained using the pattern clusters in addition to annotated training data. The operating data is also used to cluster the signals in the operating data into signal clusters. Monitored data samples are then classified by evaluating confidence vectors that include substitutions of signals contained in the training data by signals in the same signal clusters as the signals contained in the training data.

The invention belongs to the technical field of systems measuring a direction or a deviation from a predetermined direction and discloses a method and system for estimating a two-dimensional angle-of-arrival of a satellite signal. The method comprises a step of performing initial suppression on non-Gaussian clutter by using fractional low-order circular correlation, a step of performing secondarysuppression on the non-Gaussian clutter by using an improved zero-order statistical magnitude, a step of obtaining a signal subspace by an optimization method and obtaining a three-dimensional map oftwo-dimensional angle-of-arrival estimation by a multi-signal clustering method, and a step of performing maximum value search on the three-dimensional map to obtain an estimated value of the two-dimensional angle-of-arrival in a non-Gaussian clutter scene. When a generalized signal-to-noise ratio is greater than 10 dB, a root mean square error of azimuth and pitching angle estimation is smaller than 10<-1>. That means that the system and the method have good estimation performance for the two-dimensional angle-of-arrival of the satellite signal in the non-Gaussian clutter scene under a low SNR condition, and a moving target can be tracked and located through satellite radiation source signals by using the method and the system.

The invention relates to a pulse signal cluster sorting method based on class merging. The method aims to overcome the defects that the class number accuracy in the existing cluster result is low and the cluster number is inconsistent with the real signal number after class merging. The pulse signal cluster sorting method based on class merging specifically comprises the following steps: 1, determining initial cluster centers and sorting distances; 2, obtaining new cluster centers; 3, calculating whether the new cluster centers satisfy signal features; and 4, merging the cluster centers satisfying the signal features, thereby accomplishing signal cluster sorting based on class merging. The method is applied in the field of signal processing.

The invention discloses a monitoring alarm signal clustering method taking a power grid event as a center, and the method comprises the following steps: 1, extracting feature information based on a standard signal in a monitoring information specification, and forming a standard signal feature point mapping model; 2, based on expert experience of a power system, constructing a power grid event model according to logic of manually analyzing power grid events; 3, based on historical power grid event analysis report, historical monitoring data and measurement data, performing event model deduction and improvement; 4, performing structural analysis on the monitoring alarm information, and matching a standard signal; 5, performing event analysis in combination with the power grid topology and the operation mode, and performing multi-dimensional cross confirmation in combination with the telemetry information; and 6, clustering the monitoring alarm signals related to the events based on an event analysis result. According to the invention, the processing confirmation time of monitoring alarm signals is effectively reduced, the working pressure of operators is reduced, and the monitoringmanagement level of a power grid is improved.

The invention discloses a micro-vibration signal clustering method and device based on improved K-means, and the method comprises the following steps: S1, transmitting a micro-vibration signal to a K-means control machine, and generating sample data and an initial clustering center; s2, calculating a DTW distance between the sample data and an initial clustering center; s3, comparing DTW distances, carrying out first clustering after labels are marked, transmitting a result obtained through the first clustering to a DBA updater, and obtaining a new clustering center; s4, comparing whether theinitial clustering center is equal to the updated clustering center or not, if yes, executing the step S5, and if not, executing the step S6; s5, outputting a clustering result; and S6, returning to the step S3-S4, and carrying out next clustering. The method and device are more suitable for micro-seismic signal clustering analysis, waveform characteristics are reserved to the maximum extent, andthe method and device are high in processing efficiency and good in reliability.

Provided is a concrete corrosion acoustic emission signal data analysis processing system. The system comprises an acoustic emission signal acquisition instrument, a computer is installed with a software platform based on matlab, the matlab-based software platform is used for performing analysis processing on acoustic emission signal data, and the matlab-based software platform comprises a sensorcalibration module used for obtaining the sensitivity of a sensor through comparison of a given signal and an acquired signal, a signal processing module used for obtaining mutual relations between amplitude, frequency and time of signals and signal clustering characteristics through Fourier transform, wavelet transform and clustering analysis, a parameter analysis module used for determining theconcrete mechanical behavior according to change characteristics of acoustic emission signal parameters with time and frequency etc., and a signal recognition module used for realizing correspondenceof acoustic emission signal characteristics and concrete damage characteristics and performing source positioning and damage recognition. According to the system, an interactive interface is adopted,the structural design is reasonable, the work efficiency is high, and the adaptability is high.

An apparatus, such as a subscriber unit or a base station within a spread spectrum communication system, provides advanced control over the time-tracking of demodulation elements when unresolvable multipath situations arise. The apparatus provides merge protection that prevents clustered demodulation elements from contracting beyond a minimum time span. In addition, the apparatus provides a master / slave feature for synchronizing the time-tracking of the demodulation elements when clustered around a multipath signal.

The invention belongs to the field of electrical equipment partial discharge signal state detection, and discloses a partial discharge signal clustering method under multi-source discharge and interference superposition conditions, which comprises the following steps of 1 inputting sample data to be clustered; 2 setting an initial clustering number c=1, and inputting the initial clustering number;3 under the initial clustering number, selecting an optimal clustering algorithm from the four clustering algorithms of a fuzzy C-means clustering algorithm, a Gaussian mixture model clustering algorithm, a GK fuzzy clustering algorithm and a fuzzy maximum likelihood clustering algorithm; 4 continuously setting the clustering number as c+1, repeating the step 2 and the step 3 until c+1 is greaterthan 5, and entering the step 5; and 5 selecting the clustering number and the clustering method with the optimal clustering effectiveness index, analyzing the sample data according to the optimal clustering number and the clustering method, and outputting a clustering result to realize the automatic optimization of the partial discharge signal clustering algorithm and the clustering number.

The invention relates to a tone quality processor based on room impulse response measurement. The tone quality processor includes a core processor, a plurality of digital signal processing sub-circuits are arranged in the core processor; the digital signal processing sub-circuit comprises a system configuration parameter loading sub-circuit, a user-defined swept-frequency signal generating and sending sub-circuit, a multi-channel signal clustering and merging sub-circuit, a system pulse response calculation sub-circuit, a real-time sound pressure level monitoring sub-circuit, a listening sensecompensation effect preset sub-circuit, a target pulse response design sub-circuit, an equalization inverse pulse generation sub-circuit, a virtual pulse response synthesis sub-circuit, a single-channel real-time convolution sub-circuit, a sound box-listening point distribution matrix sub-circuit and an input / output distribution matrix sub-circuit. Self-adaptive tone quality balance and sound field effect compensation of different sound field systems are realized, so that great convenience is brought to a sound equipment engineering debugging process and flexible control of a user on sound field control.

The invention discloses a method for performing clustering analysis and identification on substances by adopting Raman spectrums, and relates to a substance identification method. The method comprises the steps of (1) acquiring the Raman spectrums of samples through a Raman spectrometer, and forming a sample set consisting of p samples; (2) calculating difference coefficients among all the samples, and establishing a difference coefficient matrix; (3) removing isolated samples according to the difference coefficients among the samples, and performing re-arrangement to form a new difference coefficient matrix; (4) sequentially clustering various samples of the sample set from the new difference coefficient matrix; (5) repeating the step (4) until all elements are distinguished; and (6) re-arranging an original sample sequence according to a clustering sequence, and re-drawing the difference coefficient matrix for enabling various samples to be accurately distinguished into corresponding blocks. According to the method, Raman signal clustering under high background interference is finished by utilizing the difference coefficients of the spectrums; a clustering method is unsupervised; and the method is simple, convenient and low in cost, meets exploratory detection and field actual application demands, and can be widely applied to the fields of drugs, foods, chemicals and the like.

The embodiment of the invention provides a statement block packaging method and device, electronic equipment and a storage medium, which are applied to the technical field of electronic design automation, and the packaging method comprises the following steps: determining a plurality of statement blocks to be packaged according to a segmentation boundary obtained by an RTL segmentation tool; traversing each to-be-packaged statement block, wherein a module where the current to-be-packaged statement block is located is used as a current processing module; and scanning the current processing module, determining whether the current processing module contains a black box signal, and carrying out packaging processing according to the black box signal clustering sub-module and the statement block. The packaging processing is performed by taking the black box signal as the link, so that the RTL segmentation tool is suitable for packaging the submodules and statement blocks under different grammar rules, the processing efficiency is improved, and rapid layout planning is realized.

The invention discloses a radiation source signal clustering sorting method based on radar pulse aliasing degree judgment. Firstly, a data acquisition platform is built, different kinds of radar signals are selected to form to-be-sorted data, and different pulse signal data in a PDW (pulse description word) of the to-be-sorted data are acquired; then, a radar signal pulse aliasing degree judgmentrule is formulated to judge the aliasing degree of the pulse signal data, and the pulse signal data with the aliasing degree lower than a set threshold are selected as data to be clustered; and finally, based on an RF-PW-characterized density clustering algorithm, clustering sorting is carried out on the to-be-clustered data to complete clustering sorting of the to-be-sorted data. According to themethod, the signals of various aliasing states are subjected to aliasing state determination, the low-aliasing-state or non-aliased pulse signals are selected for big data clustering sorting, and then big data clustering sorting is carried out on the remaining parts, so that the sorting and recognition success rate can be increased, and meanwhile, the sorting time consumption can be reduced.

The invention relates to the electromagnetic environment signal monitoring field and particularly relates to a signal clustering method. The method comprises steps that firstly, a pulse description word PDWi is read by a receiver, according to the orientation Doai and the carrier frequency Rfi of the pulse descriptor PDWi, the PDWi index address is calculated; secondly, the content stored in the PDWi index address is determined, if the lower-adjacent index address of the orientation Doai has no channel number mark, the upper-adjacent index address of the carrier frequency Rfi of the same orientation Doai is detected; the content stored in the index address is determined, whether the index address has a channel number mark is determined, if the upper-adjacent index address of the carrier frequency Rfi has no channel number mark, the lower-adjacent index address of the carrier frequency Rfi of the same orientation Doai is detected. An index address hash table method is provided, detectedsignals can be timely and rapidly clustered, 100% signal acquisition is guaranteed, moreover, the upper and lower-adjacent relationship of the orientation Doai and the carrier frequency Rfi are further applied, a carrier of the signals is not split due to the motion or the frequency agility of radar signals, and correct signal sorting can be guaranteed for subsequent signal processing.

The invention belongs to the field of power electrical equipment partial discharge signal monitoring, and provides a grid-based partial discharge signal clustering analysis method, relates to a data mining technology and a digital signal processing technology. The method comprises the following steps: firstly, acquiring an original partial discharge signal through an acquisition system to obtain an effective signal characteristic quantity; performing normalization processing on the original effective characteristic quantity through preprocessing; dividing grid units by utilizing the attributes of the data set; mapping the data into a grid through a certain index relation; estimating the local density of the grid by using the advantages of a Gaussian kernel function; self-deciding the clustering number and the clustering center point in the form of a decision diagram by using the idea of relative distance; distributing non-clustering center data according to the clustering center point; and finally, marking noise points and outliers according to the relative distance. The grid-based partial discharge signal clustering analysis method provided by the invention has a series of advantages of easy thought realization, easy integration, short development period, high real-time performance, less clustering result human intervention and the like.

The present invention relates to a wavelet decomposition-based high frequency partial discharge adaptive filtering and clustering method and device. The filtering and clustering method comprises the steps of using a wavelet decomposition module to carry out the wavelet decomposition on the acquired original waveforms to obtain the high-frequency components and the low-frequency components, and transmitting the high-frequency components and the low-frequency components to a component signal to noise ratio calculation module; using the component signal to noise ratio calculation module to receive the high-frequency components and the low-frequency components transmitted by the wavelet decomposition module, separately calculating the signal to noise ratios of the high-frequency components andthe low-frequency components separately, and transmitting to a filtering module; using the filtering module to receive the signal to noise ratios of the high-frequency components and the low-frequency components transmitted by the component signal to noise ratio calculation module, extracting the component having the highest signal to noise ratio, taking the waveform of the component having the highest signal to noise ratio as an analysis waveform, and transmitting the analysis waveform to a signal clustering module; using the signal clustering module to receive the analysis waveform transmitted by the filtering module, and recording a decomposition coefficient of the analysis waveform, thereby classifying the original waveforms having the same decomposition coefficients into a category.

The invention relates to the technical field of electromagnetic target intelligent clustering, and discloses an electromagnetic target intelligent clustering method based on bispectral characteristics. The method comprises the steps: performing the preprocessing of pulse-by-pulse DC removal, noise reduction and energy normalization by utilizing received radiation source intermediate frequency data, and then extracting bispectrum (three-order cumulant) characteristics of pre-monopulse data one by one; performing clustering calculation by utilizing a Kmeans method by calculating the Euclidean distance of each monopulse bispectrum sequence, and completing the intelligent clustering of targets is completed. According to the invention, the characteristic that the bispectral characteristic of the pulse signal is not easily polluted by Gaussian white noise is utilized; the method directly starts from the collected intermediate frequency data, does not need to extract the intra-pulse and inter-pulse parameters of the target, is not sensitive to the similarity of the radiation source parameters, and can improve the signal clustering capability of the radiation source with similar parametersin a complex environment.

The invention relates to an unsupervised radar signal sorting method based on deep clustering. The method comprises the following steps: carrying out the signal modulation format sorting of a to-be-sorted time sequence signal through an unsupervised sorting model, and obtaining a radar signal sorting result. The unsupervised sorting model is obtained through the combination of a deep self-coding network layer and a time sequence signal clustering layer, the deep self-encoding network layer comprises an encoding layer and a decoding layer, the encoding layer compresses an input time sequence signal into a more compact potential feature representation vector in a dimensionality reduction manner, and the time sequence signal clustering layer performs unsupervised clustering analysis on the potential feature representation vector to obtain the radar signal sorting result; and a reconstruction loss function and a KL contrast divergence loss function are used as a total cost function of the model, network weight parameters and a clustering center are reversely updated by minimizing the total cost function, and joint optimization training is performed on the unsupervised sorting model. The whole training process of the model is unsupervised, and efficient and accurate sorting of radar signals can be realized.

The embodiment of the invention provides a light control method and device based on illuminance clustering and a support vector machine. The method comprises the following steps: acquiring illuminance, time, sound intensity and infrared signals, clustering the illuminance by adopting a K-means algorithm, and mapping the time, the sound intensity, the infrared signals and the clustered illuminanceto a high-dimensional space by adopting a Gaussian kernel function to obtain a high-dimensional space feature vector; and taking the high-dimensional spatial feature vector as an input, and training the model by using a sequence minimum optimization algorithm SMO in a support vector machine to obtain an effective support vector machine model; and predicting the high-dimensional spatial data acquired in real time by adopting an effective support vector machine model, and sending a turn-on or turn-off instruction to the lamp bank according to a prediction result. According to the invention, thelight can be turned on or turned off in proper occasions and time, manual inspection on the turning on or turning off of the light is not needed, the economic cost is saved, and the application rangeis wider.

The invention discloses a radio signal clustering method based on deep learning, and the method comprises the steps: S1, adjusting the sizes of sample data of a second data set and a first data set to be consistent, and dividing the sample data of the two data sets into a plurality of batches containing the same number of sample data; S2, constructing a depth model, and inputting the adjusted second data set batch sample data into the depth model for pre-training until the model training is stable; S3, performing clustering training; and S4, outputting a clustering result. The invention further comprises a radio signal clustering system based on deep learning. According to the method, a neural network combining one-dimensional convolution and two-dimensional convolution is established, and a modulation signal data set is used for pre-training a depth model, so that the depth model tends to extract features related to modulation types, and during subsequent clustering, the model is guided to cluster samples according to the modulation types. The method has good universality in the aspect of signal clustering.

The invention discloses a structure variation detection method and device and a storage medium. The method comprises the steps of obtaining a comparison file, extracting reads in an interval from the comparison file according to a set length, and dividing abnormal reads into DP signals, SR signals and SU signals; the DP signals are clustered, each cluster serves as a structural variation candidate, and local assembly and re-comparison are carried out on each cluster; finding an embedding comparison from the SR signals, and carrying out assembly and re-comparison; and performing fusion breakpoint left and right side mutation depth calculation and structure variation type identification on two re-comparison results. According to the method, by means of DP signal clustering and assembly re-comparison, false positive signals in clusters are reduced; and the SR signal analysis is used for supplementing, so that the detection rate and the precision of the whole result are higher. According to the method, structural variations such as deletion, inversion, repetition, translocation in chromosomes, translocation between chromosomes and the like can be recognized, and micro homologous sequences and short template sequences near breakpoints are provided for output.

The invention relates to a 35kV power line fault traveling wave extraction method based on affinity propagation clustering, and belongs to the technical field of 35kV power line fault distance measurement. The method comprises the steps of data preprocessing, fault traveling wave signal characteristic parameter extraction, affinity propagation clustering calculation and the like. According to themethod, the influence of interference signals generated by power noise on a 35kV power line fault accurate positioning system is reduced, switching value signals do not need to be extracted, clustering calculation is conducted on traveling wave signals through affinity propagation clustering, and misjudgment generated by interference of the interference traveling wave signals is eliminated according to the clustering characteristics of the interference traveling wave signals. Meanwhile, the method enables the 35kV power line fault accurate positioning system to identify lightning stroke multi-source discharge points, and is easy to popularize and apply.

The invention relates to a PPG signal clustering center acquisition method, a PPG signal processing method and a PPG signal processing device. The PPG signal clustering center acquisition method comprises the following steps: S11, acquiring and preprocessing a sample PPG signal to obtain a second pulse; s12, performing fast Fourier transform on the second pulse, extracting a frequency component, and normalizing characteristic parameters of the frequency component to obtain a third pulse; s13, clustering the third pulse to obtain an initial clustering cluster and an initial clustering center; s14, constructing an MAE similarity matrix according to the normalized parameter MAE values of the two initial clustering centers; s15, acquiring an MAE value of a second minimum value of row vectors in the MAE similarity matrix, when two MAE values which are smaller than a first preset value and correspond to the same initial clustering center exist, combining the initial clustering cluster corresponding to the MAE value as a new initial clustering cluster, and executing S14 by taking a mean pulse of the initial clustering center corresponding to the MAE value as an initial clustering center of the new initial clustering cluster, otherwise, taking the initial clustering center as the target clustering center.

There is provided methods and device for imaging objects comprising unsupervised classifying and data analyzing of the object to detect and identify the structure of the object and further display the object's structure / underlying structure, for example the arrangement of and relations between the parts or elements of the object by using a location module configured to record the physical location of an antenna array.

The invention discloses a method for clustering seismic pre-stack signals based on RBM. The clustering method of the invention aims at the high-dimensional characteristics of seismic pre-stack signals, and introduces a dimensionality reduction method based on a depth-restricted Boltzmann machine network. Restricted Boltzmann machine has good abstract feature extraction ability, and the output after dimensionality reduction can well express the original signal; in view of the characteristics that seismic signals are often noisy and difficult to remove, a fuzzy self-organizing neural network clustering algorithm is introduced , so that the classification of seismic facies is expressed by the degree of membership, the information provided by the fuzzy classification results is more abundant and reasonable, the clustering results are more stable, and the classification results can be further explored. The method of the invention has relatively low computational complexity and can be extended to large-scale seismic pre-stack data.

A differential signal transmission device transmits N differential signal pairs from a differential signal generator to a number of receiving terminals. The N differential signal pairs include N positive signals and N negative signals. The N positive signals are clustered at a first positive clustering point. The first positive clustering point is connected to a second positive clustering point via a first matching resistor. The second positive clustering point is grounded via a first grounding resistor, and outputs a number of positive signals to the number of receiving terminals respectively. The N negative signals are clustered at a first negative clustering point. The first negative clustering point is connected to a second negative clustering point via a second matching resistor. The second negative signal clustering point is grounded via a second grounding resistor, and outputs a number of negative signals to the number of receiving terminals respectively.

The invention discloses a radiation source signal clustering and sorting method based on radar pulse aliasing degree determination. First, a data acquisition platform is built, and different types of radar signals are selected to form data to be sorted, and PDW pulses of the data to be sorted are collected. Describe the different pulse signal data in the word; then formulate the radar signal pulse aliasing degree judgment rule to judge the aliasing degree of the pulse signal data, and select the pulse signal data whose aliasing degree is lower than the set threshold as the data to be clustered; finally The density clustering algorithm based on RF-PW is used to cluster and sort the data to be clustered, and complete the clustering and sorting of the data to be sorted. The present invention proposes to judge the aliasing state of signals in various aliasing states, select the low-aliasing state or non-aliasing pulse signal for big data clustering and sorting, and then perform big data clustering on the rest Class sorting can increase the success rate of sorting recognition and reduce the time consumption of sorting.

The invention provides a carrier synchronization method and device based on 5G high-order modulation signal clustering discrimination. The method comprises the following steps: setting a 5G communication signal basic parameter; generating two paths of baseband IQ orthogonal signals xI(t) and xQ(t) with frequency phase offset under set parameters based on MATLAB; based on a clustering discrimination method, discriminating two paths of input IQ orthogonal signals in sequence; performing frequency phase offset correction on the discriminated signals through a frequency-locked / phase-locked loop to obtain signals xIC(t) and xQC(t); and sampling and judging the xIC(t) and the xQC(t), and calculating the bit error rate BER. Based on a frequency-locked loop (FLL) and a phase-locked loop (PLL) of a clustering discrimination method, the frequency offset and the phase offset of a 5G high-order modulation baseband signal data set are corrected, the constellation diagrams of baseband signals before and after correction are compared, carrier synchronization is achieved, on the premise that certain precision is guaranteed, the calculation efficiency of carrier synchronization in 5G high-order modulation signals is improved, and engineering application is accelerated.

The invention belongs to the technical field of communication, and specifically relates to an OFDM system signal sending and receiving method. The main method of the invention comprises the following steps: on a sending end: performing identifier adding, IFFT and cyclic prefix insertion on a sending symbol in sequence, and then outputting an OFDM signal; and on a receiving end: performing cyclic prefix removal and FFT on the received OFDM signal, clustering subcarrier frequency domain signals, and distinguishing the category of the clustered symbol according to the identifier, and restoring the sending symbol. The OFDM system signal sending and receiving method provided by the invention has the beneficial effects that, compared with the traditional technology, the detection performance of a receiver can be greatly improved by the method provided by the invention, and no channel estimation needs to be carried out on the receiving end.