67 results about "Sequence segmentation" patented technology

The invention discloses an automatic segmentation method for a lung-area CT sequence. The method is characterized by comprising the steps of 1) an image I of the lung-area CT sequence is input; 2) the image I is segmented via interactive region growth; 3) according to a segmentation result, an initial contour is obtained, and seed-point coordinates of adjacent images are calculated; 4) based on the seed-point coordinates of the image I, a present image II in the sequence is segmented via the interactive region growth in step 2; and 5) step 2, 3, 4 are repeated to determine whether all images in the CT sequence are segmented, and if no, the step 3 is turned to. The automatic segmentation method automatically calculates mapping of the seed-point areas by combining the image context object characteristic continuity of the sequence to realize sequence segmentation, thereby obtaining complete three-dimensional area data of the lung, and providing basis for VOI extraction and classification of a suspected lung tubercle for a CAD system.

The invention discloses a lip language recognition method and device, computer equipment and a storage medium. The lip language recognition method comprises the following steps: carrying out standardization processing on a frame rate of an obtained original video, and separating the obtained standard video to obtain an effective audio stream and an effective video stream; tracking a human face in the effective video stream by using a human face recognition algorithm, extracting a mouth area in the human face, and obtaining a frame lip action video so as to obtain a lip image sequence; segmenting the lip image sequence by adopting a sequence segmentation rule to obtain a segmented image sequence; sequentially inputting the segmented image sequences corresponding to the lip image sequences into a lip image recognition model for recognition, and obtaining a lip image recognition result; inputting the effective audio stream into a speech recognition model to obtain a speech recognition result; and when the similarity between the lip image recognition result and the original video reaches a preset value, taking the lip image recognition result as a lip language recognition result of the original video so as to ensure the accuracy of the lip image recognition result.

The invention discloses a stomach computed tomography (CT) sequence image segmentation method based on interactive region growth, which mainly solves the problems that in the prior art, CT sequence segmentation speed is slow, and poor segmentation is easy to occur. The method includes: firstly, a seed point is selected manually in a target area to be segmented in a first image, the interactive region growth is used for performing segmentation, a center of a segmentation result and eight neighborhoods of the center are projected into a next CT image to serve as seed points, the interactive region growth is continuously used for performing segmentation to obtain the target area of the current image, and the segmentation result of the previous image is projected into a next image repeatedly to serve as a seed point to be segmented continuously until segmentation of a whole sequence is completed. Compared with a traditional serial region growth, the stomach CT sequence image segmentation method based on the interactive region growth has the advantages of being rapid in speed, good in effect and the like, can be used for segmenting stomach CT sequence images, and can well segment target areas which may occur in stomach lymph gland in the sequence.

A cell search timing synchronization method includes: performing cross-correlation operation for main synchronous signals kept by UE (user equipment) and received signals so as to obtain correlation sequences; segmenting the correlation sequences to obtain sequence segments corresponding to the correlation sequences; and computing a peak-to-average value of the sequence segment corresponding to each correlation sequence, and obtaining a timing synchronization position according to the maximum peak-to-average value of the correlation sequences. After application of the cell search timing synchronization method in the embodiment, cell search is accurately timed and completed in the presence of strong uplink interference signals.

The invention discloses a wind power unit commitment contained modeling method considering predication error timing sequence distribution. A predication error segmentation fitting method under a timing sequence based on error characteristic analysis is proposed; fitting is performed based on t location-scale distribution to reduce a heavy-tail effect and improve the fitting precision, and the fitting method can be matched with a UC decision in timing sequence; next, a dual-quantile type UC decision model capable of taking conventional cost, extra standby cost and risk cost into considerations is established; the selection of confidence levels can be balanced through the restrictive relations among different costs; the standby classification can be guided by the division of the different confidence levels; the error timing sequence segmentation distribution is adapted by time varying confidence levels, so that the model is more economical, targeted and applicable; and finally, an improved hybrid particle swarm algorithm with a heuristic searching principle is adopted to solve a multi-variable mixed integer programming model of a text, and the effectiveness of the method is verified by example results.

The invention relates to a management line loss abnormity identification method based on data mining. The technical points comprise: step 1, carrying out sub-sequence segmentation on preprocessed management line loss time sequence data by adopting a sliding window method; step 2, constructing a time sequence prediction model based on a neural network, obtaining a predicted value of a management line loss sub-sequence, and judging the sub-sequence of which the difference range between the predicted value and an actual measurement value is greater than a preset threshold value as an abnormal sub-sequence; step 3, extracting characteristic variables of the abnormal sub-sequences, establishing a management line loss characteristic sample set, and clustering by adopting three different algorithms; and step 4, performing cluster matching on the three clustering results, obtaining a final clustering result by adopting a majority voting clustering integration method, and comparing the difference between the number of objects in the cluster and a preset threshold value to obtain a specific classification condition of the management line loss abnormal sub-sequence. The abnormal condition ofline loss can be quickly and accurately identified and managed, and better stability and practicability are achieved.

The invention discloses a polar code encoding method, especially relates to a polar code encoding method of a sectional-cascaded Hash sequence, and belongs to the field of the channel encoding. The encoding method comprises the following steps: performing Hash check on information bit sequence sections, and orderly cascading the information bit sequences with checked bit sequences to construct a new bit sequence to perform the polar code encoding. The core of the method is that the check accuracy is improved by using the independent random feature of the Hash sequence, the probability of generating the same check bit by different information is lowered; and meanwhile, the sectional checking method not only can be used for checking the information bit, but also can be used for checking the former check bit.

The invention discloses a phonetic segmentation method and device for speech synthesis, wherein the method comprises: acquiring corpus text and converting it into a pinyin sequence, wherein the pinyin sequence comprises a plurality of phones, each having multiple states; segmenting speech data corresponding to the corpus text into multiple speech frames, and acquiring acoustic features of the speech frames; with respect to each state, clustering the speech frames according to the acoustic features, and generating multiple nodes corresponding to the state; calculating an optimal path corresponding to the pinyin sequence based on a dynamic programming algorithm and a two-dimensional state network, and segmenting the pinyin sequence according to the optimal path. The phonetic segmentation method and device for speech synthesis in embodiments of the invention enables the accuracy of pinyin sequence segmentation, thus improving the reliability of a speech synthesis acoustic model and enable speech in text-to-speech conversion to be smoother and more natural.

The invention provides a road intersection control period dividing method and system based on multidimensional time sequence segmentation, and the method comprises the steps: configuring a road intersection stationary information table; taking parameters or indexes for reflecting the traffic demand features as the period dividing basis, and obtaining a data time sequence; gathering the traffic flow data at different spatial levels according to the road intersection stationary information table; determining a road intersection traffic flow time sequence dimension d through road intersection canalization conditions; carrying out the time-dimension gathering of the traffic flow detection data; constructing a T*d basis matrix C; employing a dynamic planning and correction algorithm of non-parameter multi-change-point analysis for determining a dividing number and a specific dividing scheme of the basis matrix C. According to the invention, an automatic period dividing mechanism reduces manual intervention, and the traffic demand features are taken as the dividing basis, thereby reducing the technical requirements for a signal configuration link in an application, and improving the period dividing reliability and stability.

The invention provides a segment parallel coding method of a feedforward convolutional code. The method is characterized in that: carrying out parallel coding on input information bit sequence by a plurality of serial feedforward convolution encoders, carrying out segmentation on an input information bit sequence to be coded, respectively applying a serial feedforward convolution encoder to code each segment of the information bit sequence, wherein, each segment of the information bit sequence is named as a block, in segmenting, overlapping bits, or after coding, inputting full 0 bits for supplement, generating overlap, according to an overlapped part, connecting code results of each serial feedforward convolution encoder in series, and taking the results as final output of parallel feedforward convolution coding. According to the invention, based on information sequence blocking, a characteristic of a feedforward convolution encoder is fully utilized, resource occupation and power consumption are reduced, and coding throughput is raised.

The invention discloses an association rule mining method for multivariate time series monitoring data. Objects come from a complex engineering system. The method comprises the following steps: firstly, determining an association relationship of time sequences through a time sequence correlation measurement method and clustering the time sequences of the association relationship; secondly, determining an abnormal value according to the relative change rate of the time sequence; determining an abnormal trend by setting an observation window; determining abnormal distribution through local datadensity, determining abrupt change points through linear fitting, finally, segmenting the sequence through trend, carrying out pattern representation on the time sequence through the slope of a fitting straight line and trend duration, setting the format of a time characteristic rule, and completing association rule mining through a frequent pattern.

The invention relates to word segmentation technology and provides a word input method and a word input device aiming at the defects of large amount of computation and low identification accuracy of ambiguous words in the existing word segmentation methods. The word input method comprises the steps of receiving the input word-building element sequence and searching and displaying at least a word string matched with the word-building element sequence; and receiving the input word string selection command and generating and outputting corresponding word segmentation packages based on the selected word string in the displayed word strings. The invention also provides a word input system. The method and system provided by the invention can automatically finish the word sequence segmentation in the input process; therefore, the technical scheme provided by the invention can greatly reduce the amount of computation of existing word segmentation operation and improve the word segmentation efficiency. In addition, the real segmentation intention of the user can be accurately reflected according to the segmentation on the word sequence conducted by the word segmentation packages, thus greatly improving the word segmentation accuracy of the word sequence.

The invention discloses a trill identification method and a trill identification device. The trill identification method comprises steps that target audio data corresponding to a target song comprising at least one identification note is acquired; the audio data fragment of the target audio data corresponding to the identification note is extracted, and the first fundamental frequency of the audio data fragment is extracted, and then a corresponding first note value sequence is acquired; the first note value sequence is divided into at least two note value sequence fragments according to at least one preset periodic value, and by aiming at every preset periodic value, a note distance between the two adjacent note value sequence fragments of the at least two note value sequence fragments is calculated, and is used as the periodic note distance corresponding to the periodic value; the minimum value of the periodic note distance is determined, and a target period corresponding to the minimum value is acquired; when the target period is smaller than a preset periodic threshold value, the target audio data is determined to be trill audio data. By adopting the trill identification method and the trill identification device, the identification of the trills of the audio data is realized.

The present invention discloses an MIC based random number generator segmented parallelizing method, so as to perform parallelizing by a method of segmenting a periodic sequence and splice random numbers generated by threads to form a final sequence. Compared with a CPU single thread, a speedup ratio under an MIC platform has a significant advantage.

The invention relates to a non-linear wind speed simulation method considering divided period. The method involves two parts of wind speed time sequence division and non-linear time sequence fitting and comprises the steps of firstly, deriving to obtain a division module according to a maximum posterior probability expression, and simultaneously performing division on a non-stationary time sequence by considering distribution characteristic of wind speed sequence, wherein the division module can take actual experience of engineering as wind speed prior information, each segment of sequence after division is a stable sequence, so that the hypothesis that the sequence is required to be stable by a non-linear SIAVAR single index which can be added to a vector autoregression model is solved; secondly, respectively simulating each segment of sequence by applying a non-linear SIAVAR model, wherein non-linear simulation adopts a spline function as global approximation of a non-parameter function in the non-linear model, and half-parameter estimation is performed by Backfitting method; and finally, completing wind speed sequence fitting. By the method, the non-linear characteristic of a wind speed can be simulated very well.

The invention discloses a method for segmenting handwritten characters of a financial bill small writing amount. The method is characterized by comprising the following steps: 01, searching an optimalsegmentation threshold for an input grayscale image by utilizing a K-means algorithm; 02, performing transverse and longitudinal projection segmentation on a character sample to be identified by using a histogram method to obtain a preliminary character segmentation sequence; and 03, using the character sequence segmentation sequence to calculate the central probability, namely the correlation coefficient, between the two characters through a CRAFT algorithm, if the correlation coefficient between the number amount characters is large and the correlation coefficient between the grid verticallines and the handwritten characters to be small, excluding grid lines, screening the optimal character segmentation sequence through the CRAFT algorithm, and outputting the result.

The invention provides a method and device for predicting epidemic situation data in a segmented mode, a computer readable medium and electronic equipment, and relates to the technical field of medical data processing. The method comprises the following steps: acquiring an effective regeneration number corresponding to each unit time of a target disease in N continuous unit periods to obtain a first sequence containing N effective regeneration numbers, N being an integer greater than 2; segmenting the first sequence to obtain a plurality of sub-sequences, and determining a fitting function corresponding to each sub-sequence according to the segmented sub-sequences; and predicting epidemic data of the target disease based on the plurality of fitting functions. According to the technical scheme, the epidemic situation data can be subjected to related prediction on the basis of the effective regeneration number, so that the method has more practical application significance; meanwhile, the effective regeneration number is fitted in a segmented mode, prediction is achieved on the basis of multiple fitting functions, the prediction accuracy can be improved, and the epidemic situation data prediction accuracy is improved.

The invention discloses an HLA sequencing peak graph identification method, which comprises the following steps of: (1) constructing a peak graph identification model: (1) collecting machine peak graph data under the existing HLA, and (2) pre-processing to complete information extraction, original sequence comparison, sequence segmentation and dislocation repair work of a binary ab1 file; 3, feature extraction; (4) collecting a large amount of artificially recognized data, training the data by using a random forest algorithm, and constructing a peak graph recognition model; (2) carrying out base recognition on the to-be-detected HLA first-generation sequencing original off-machine data by utilizing a peak diagram recognition model; (3) arranging the recognized base sequence, and reassembling the single-chain and double-chain part sequences; and (4) outputting an identification result. According to the recognition method, the peak diagram sequence information can be accurately obtained,the overall accuracy is 99.5% or above, and the working efficiency of HLA data interpretation personnel is greatly improved.

The invention discloses a coding method and device, electronic equipment and a storage medium, which are used for solving the problem of inflexible segmentation of a sequence to be coded in the priorart. The method comprises the following steps: determining a target number of sequence segments to be coded according to the length and transmission code rate of a sequence to be coded; According to the target number, carrying out segmentation processing on the to-be-coded sequence; And encoding each subsequence subjected to the segmentation processing, and cascading after encoding. In the embodiment of the invention, the target number of the sequence segments to be coded is determined according to the length and the transmission code rate of the sequence to be coded, and the sequence to be coded is segmented according to the target number. And the sequence to be coded is flexibly segmented, so that the coding performance is improved.

The invention discloses a rainfall prediction method and system based on an artificial intelligence algorithm and a knowledge graph, and belongs to the technical field of weather forecast, and the method comprises the following steps: constructing a multi-modal data container, inputting different types of meteorological data into the multi-modal data container according to the structural characteristics of the data, and obtaining multi-modal data; performing space-time alignment, data cleaning and preprocessing and data sequence segmentation on the multi-modal data; constructing a knowledge graph about rainfall prediction; and creating a multi-modal rainfall prediction model based on an artificial intelligence algorithm, and carrying out intelligent correction on predicted rainfall data. According to the method, different types of multi-modal meteorological data are fully utilized, the rainfall prediction precision is improved through the multi-modal rainfall prediction method, a rainfall prediction knowledge graph is constructed, a model prediction result can be intelligently corrected, the uncertainty of an artificial intelligence algorithm is reduced, and the rainfall prediction accuracy is improved. And the accuracy and reliability of model prediction are improved.

The embodiment of the invention provides a character recognition method and device, electronic equipment and a readable storage medium, and aims to improve the recognition accuracy of character recognition. The method comprises the steps of obtaining N character sequence segmentation images of a to-be-recognized image, wherein the pixel value of each pixel point in the nth character sequence segmentation image in the N character sequence segmentation images represents the possibility that a character corresponding to the pixel point is the nth character; obtaining M character category segmentation images of the to-be-identified image, wherein the pixel value of each pixel point in the mth character category segmentation image in the M character category segmentation images represents the possibility that the pixel point belongs to the mth character; and determining a character sequence and a character category of each character in the to-be-identified image according to the pixel valueof each pixel point in each character sequence segmentation image and the pixel value of each pixel point in each character category segmentation image.

The invention provides a rib instance segmentation, counting and positioning method and system, and the method comprises the following steps: carrying out the semantic segmentation of ribs in a chest CT, and obtaining a rib semantic segmentation binary mask; collecting a manual label corresponding to the rib semantic segmentation binary mask to obtain a rib instance segmentation mask; processing the rib semantic segmentation binary mask and the rib instance segmentation mask, and making a sequence segmentation training sample set; establishing a deep convolutional neural network model; training the deep convolutional neural network model by using the training sample set to obtain a deep convolutional neural network capable of predicting the topmost ribs; inputting the rib semantic segmentation binary mask into the trained deep convolutional neural network, obtaining a prediction mask of the topmost layer of ribs from top to bottom in sequence, and further obtaining a predicted rib instance segmentation mask. According to the method, explicit modeling is carried out on the rib counting problem, the ribs are segmented one by one from the lung top to the lung bottom, and rib counting and positioning are achieved.

The invention relates to an industrial control protocol analysis method, in particular to an efficient industrial control protocol analysis method based on deep learning, which comprises the following steps: acquiring flow data messages of an industrial control system from a simulation platform and an open source platform, analyzing protocol field change characteristics, and adopting an unsupervised learning method to analyze the flow data messages of the industrial control system. A voting expert algorithm (VE) performs sequence segmentation and format feature inference on protocol fields. And taking the processed field sequence features as input, building a bidirectional long-short-term memory neural network model (BiLSTM-AM) added with an attention mechanism, performing training, using softmax as a protocol field classifier, and realizing industrial protocol field classification result prediction according to a classification result. According to the invention, based on the bidirectional long-short-term memory neural network model added with the attention mechanism, a good detection result is achieved in unknown industrial protocol prediction and classification; the problem that most network protocol reverse tools in the current market cannot efficiently and accurately analyze the industrial protocol is solved.

The invention discloses a method of acquiring an rDNA ITS2 sequence for identification of dendrobium candidum, provides a standard sequence for identification of dendrobium candidum, and can effectively solve the molecular identification problem of dendrobium candidum medicinal materials. The specific steps include: (1) designing two pairs of primers to conduct segmented PCR amplification on an ITS2 sequence; (2) sequencing the amplification results, and connecting the results together according to overlapping regions, thus obtaining an ITS2 sequence; and (3) performing comparison with the standard sequence SEQ ID No.1, and if the sequence similarity is greater than or equal to 99%, identifying the sample as Dendrobium officinale. The method provided by the invention can realize rapid and accurate identification of dendrobium candidum herbs, original plants, seeds and seedlings, etc.

A method for accelerating time series data base queries includes segmenting an original time series of signal values into non-overlapping chunks, where a time-scale for each of the chunks is much less than the time scale of the entire time series, representing time series signal values in each chunk as a weighted superposition of atoms that are members of a shape dictionary to create a compressed time series, storing the original time series and the compressed time series into a database, determining whether a query is answerable using the compressed time series or the original time series, and whether answering the query using the compressed time series is faster. If answering the query is faster using the compressed representation, the query is executed on weight coefficients of the compressed time series to produce a query result, and the query result is translated back into an uncompressed representation.

The invention provides a transformer DGA online monitoring data processing method, which comprises the following steps: according to the characteristics of returned data, enabling online data to be equivalent to a time sequence; in the first stage, the thought of a sliding window algorithm is introduced, an improved sequence piecewise linearization algorithm is provided, sequence data is divided into a plurality of line segments represented by slopes and spans, then online monitoring data is symbolized by using improved K-means clustering, and finally the relevance among different indexes in DGA is mined by using an APRIORI algorithm, so that the DGA is subjected to online monitoring. An abnormal value existing in the data is explored; and in the second stage, according to the screened abnormal numerical sampling points, an improved particle swarm optimization support vector regression algorithm is used, the solving speed and solving diversity of the algorithm are guaranteed, key parameters in the support vector regression algorithm are optimized, and the sampling points are repaired, so that the processing of the online DGA monitoring data of the transformer is completed.

The embodiment of the invention discloses a micro-seismic signal reconstruction method and system utilizing conversion sparsity. The method comprises the steps of 101, acquiring a signal sequence S acquired according to a time sequence; 102, solving a signal comparison sequence; 103, solving the total number of sequence segments; 104, solving a signal segmentation sequence; 105, solving a controlfactor; 106, solving a conversion sparse matrix; and 107, solving a reconstructed signal sequence.

The invention relates to a polarization code decoding method, in particular to a polarization code decoding method based on a segmented Hash sequence, and belongs to the field of channel encoding. According to the polarization code decoding method, the Hash sequence is inserted in an information bit sequence in segments, the information bit sequence is checked in segments by using the Hash sequence in a decoding process, and a decoding search path is reduced or the decoding is terminated in advance by using the check result, thereby reducing the decoding complexity and reducing the decoding delay.