31 results about "Average entropy" patented technology

The invention relates to a burst traffic detection method based on dynamic threshold. By means of the method, the burst traffic in a high-speed network environment can be effectively detected on line in real time, and the problem that a fixed threshold method cannot well adapt to the dynamically changing network environment is solved. According to the technical scheme, a threshold is dynamically adjusted according to the source IP entropy value of the normal historical traffic, a sliding window mechanism is introduced, and burst traffic is detected by comparing whether the difference value between the current detection value and the average entropy value of a sliding window exceeds the dynamic threshold; the method comprises steps as follows: generation of an initial sliding window, generation of a reference value, generation of the current detection value, abnormality judgment of the current detection value, processing of normal traffic and processing of burst traffic. The detection method is rapid, simple, high in detection rate and low in false rate, the network safety is effectively guaranteed, and the detection method has high practicability and huge economic and social benefits.

The invention provides a rotor crack fault diagnosis method based on variational modal decomposition and a gray co-occurrence matrix, and belongs to the field of crack fault diagnosis. The method comprises the following steps of collecting a vibration signal; performing variational modal decomposition on the collected signal; generating a symmetric polar coordinate image for each IMF component; converting the symmetric polar coordinate images into gray images; generating the gray co-occurrence matrix by the gray images, and extracting image texture features to serve as feature parameters; selecting a feature statistics entropy of the gray co-occurrence matrix; obtaining the ith IMF component of a certain working condition; calculating average entropies of samples in corresponding states infour directions; collecting a plurality of to-be-diagnosed samples, and extracting eigenvectors; calculating Mahalanobis distances; and comparing values of d1 and d2 of to-be-detected samples, wherein the to-be-detected samples with the relatively short comprehensive distances are the states corresponding to the to-be-diagnosed samples. According to the method, the deficiency that the gray co-occurrence matrix is merely adopted for processing a rotor crack fault signal is overcome, so that the rotor crack fault diagnosis can be well completed.

The invention discloses a virus identification method comprising the following steps: obtaining RNA sequencing data of a to-be-tested sample; assembling the first portion of the sequencing data so as to obtain assemble sequences; comparing the first portion of the sequencing data with the assemble sequences so as to obtain a comparison result; determining mutation sites on the assemble sequences according to the comparison result, and determining at least one of the following assemble sequences a-c: a, at least one from the group formed by average entropy and median entropy, and mutation site proportion; b, at least one from the group formed by the average mutation rate and median mutation rate, and the mutation site proportion; c, mutation site proportion; comparing at least one from the assemble sequences a-c with the corresponding boundary, and determining the assemble sequence falling into the boundary to be from virus. The invention also discloses a virus identification device; the virus identification method and / or device can accurately predict whether an unknown sequence is a virus sequence or not without depending on the homologous sequence alignment.

The invention relates to a maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method and belongs to the image matching field. Scale-invariant?feature?transform (SIFT) operators have strong matching ability, while, the scale-invariant?feature?transform (SIFT) operators will bring a huge amount of data, and therefore, binaryzation should be performed on the scale-invariant?feature?transform (SIFT) operators, however, if unified binaryzation is performed on all the operators, data redundancy or information loss will be brought about. The maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method of the invention comprises the following steps that: binaryzation is performed on the scale-invariant?feature?transform (SIFT) operators; average entropy calculation is performed on each layer of binaryzation results, such that different numbers of binaryzation layers are selected adaptively; new binaryzation descriptors are provided; the Hamming distance is utilized to replace the Euclidean distance so as to calculate the distance between two descriptors; and the distance between the two descriptors is compared with a set threshold value. With the maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method of the invention adopted, information of original features is reserved; the amount of data storage can be greatly reduced; computational complexity can be reduced; a requirement for a real-time property can be realized better; the matching results equivalent to original scale-invariant?feature?transform (SIFT) descriptors can be obtained and are far superior to results of matching by using the unified binaryzation.

The invention discloses a fast detecting method and device for a static vehicle target. The fast detecting method includes the following steps that images are collected, wherein vehicle images are obtained; the images are preprocessed, wherein normalization is conducted on the sizes of the vehicle images, and the vehicle images are converted into a grey-scale map; entropy information is extracted, wherein pixel level traversal is conducted on the images through search boxes, and pulse iteration average entropy information of the partial images in the search boxes is extracted; rough positioning is conducted on the vehicle target, wherein the images with the pulse iteration average entropy value larger than that the target entropy value are taken as vehicle target rough positioning images; the vehicle target is determined, wherein the vehicle image with the largest pulse iteration average entropy value in the same picture area is reserved as a target vehicle search result; the vehicle target is presented in an original large image, and positioning of the vehicle target is completed. According to the fast detecting method and device for the static vehicle target, robustness, rapidness and transportability of a pulse coupling neural network feature extraction method are used, the feature of the pulse iteration average entropy is selected as the judgment basis, and the speed of detecting the static vehicle target is greatly increased.

The invention discloses a quality evaluation method based on track data. The quality evaluation method comprises the steps that S1. the track data are expressed as a tuple combination of four dimensions, and a map is obtained; S2. the track data are grouped according to an entity, a time dimension is used for sorting on groups, and each track of the track data is converted into a collection of orderly segments; S3. the map is divided into cells with the same size, and the track data are projected into the cells; and S4. the entropy of the cells is computed, and the average entropy and the weighted average entropy of all the cells are computed. According to the method, the entropy is used for effectively evaluate track data quality, data quality comparing of a plurality of track data is achieved, the feasibility for data analysis is analyzed, and data quality evaluation is easy, effective and low in cost.

The invention discloses a radiation source individual feature extraction method. The method includes the following steps: setting two-dimensional position coordinates corresponding to each cloud droplet as S (i)=(x(i), y(i)), and solving average values of all cloud droplets; solving an entropy value of each cloud droplet according to an entropy value formula; calculating an average entropy value of n corresponding cloud droplets; calculating mean square error of the entropy values, and utilizing the above result to acquire super entropy according to a super entropy calculation formula. On the basis of conventional feature extraction, a cloud cluster model is built for conventional fuzzy features acquired at a low signal-to-noise ratio, and distribution characteristics of a feature cloud cluster are depicted by utilizing basic digital features of a cloud model theory so as to realize finer signal feature extraction.

The invention discloses a recommendation system attack detection method based on time series data. The method includes the steps that a user-project preference degree data set and a frequent item set excavation technology are utilized to acquire user groups and project groups; group preference degree value proportional characteristics are calculated for the pairs of user sets and the pairs of project sets; all preference degrees of projects in the project groups form time series preference degree data according to operating time; group preference degree time interval characteristics are calculated for the pairs of user sets and the project sets; group average entropy characteristics are calculated for the user groups; for the user groups, maximum group preference degree value proportional characteristics corresponding to the user groups and maximum group preference degree time interval characteristics corresponding to the user groups are selected, and the user groups are sequentially ranked through the three characteristics, and then three ordered user group sequences are acquired; the three ordered user group sequences are integrated to acquire a wholly ordered user group sequence, and the most probably attack user groups are acquired; the most probably target project groups are acquired through the group preference degree value proportional characteristics.

A system for storing a specification document to be checked on a storage device in a computer readable format. A computer implemented method which stores a graph structure of rules for performance of checking, and creates a rule application table of whether a rule is an application subject. For each node rule of the graph structure, average entropy of whether a specification satisfies the rules is derived by searching the graph structure of rules. While performing a depth-first searching of the graph structure of rules, a priority order index is derived from the rule application table for each node rule. Amendment candidates are then displayed according to value of the priority order index.

A voice activity detection system (100) filters audio input frames (102), on a frame=by-frame basis through a gammatone filterbank (104) to generate filtered gammatone output signals (106). A signal energy calculator (108) takes the filtered gammatone output signals and generates a plurality of energy envelopes. Weighting factors are constructed (112) are applied to each of the energy envelopes thereby producing normalized weighted signal (116), in which voice regions are emphasized and noise regions are minimized. An entropy measurement (118) is taken to extract information from the normalized weighted signals (116) and generate an entropy signal (120). The entropy signal (120) is averaged and compared to an adaptive entropy threshold (122), indicative of a noise floor. Decision logic (124) is used to identifying speech and noise from the comparison of the averaged entropy signal to the adaptive entropy threshold.

The invention provides a satellite-borne remote sensing image compression method based on self-adaptive block compressed sensing, and aims to solve the problems that the data volume of a satellite-borne remote sensing image is large, if the whole image is subjected to compressed sensing, a measurement matrix in the compression process needs to occupy more storage space, and meanwhile, the reconstruction complexity is increased. According to the method, after an image is partitioned, whether the sub-blocks continue to be partitioned or not is judged according to the average entropy value of thegray level co-occurrence matrix of each sub-block; and after the partitioning is finished, different sampling rates are distributed to each sub-block according to the properties and the average entropy of each sub-block. The space occupied by the observation matrix is reduced, the blocking effect is eliminated, and the method is suitable for compression of satellite-borne remote sensing images.

The invention discloses a wavelet analysis post-processing method and device for a power transient signal. The method is to conduct the following one or more post-processing on the wavelet coefficients after the power transient signal is analyzed by wavelet, and the result is to send power Monitoring center processing: modulus maximum extraction, singularity detection, obtaining modulus maximum points and singularity; statistical processing or cluster analysis of wavelet coefficients; neural network classification, that is, wavelet coefficients or wavelet post-processed data as neural Network input, output neural network classification results; energy analysis, to obtain energy distribution coefficient and energy fluctuation coefficient; wavelet entropy calculation, to obtain wavelet energy spectrum entropy or wavelet time entropy or wavelet singular entropy or wavelet time-frequency entropy or wavelet average entropy or wavelet distance entropy. It can effectively extract the characteristics of power transient signals, which can be applied to power system transmission line traveling wave distance measurement and fault identification, power quality analysis and equipment fault diagnosis.

The invention discloses a waste power battery consistency index optimization method based on clustering and an average entropy ratio, and relates to the field of waste power battery index evaluation,which comprises the following steps: index sea selection: any index related to battery attributes can be included into an alternative index library; Primarily selecting, rejecting invalid indexes suchas data unavailability and the like; Screening is conducted, and indexes with redundant information amount larger than a threshold value are deleted through correlation analysis and clustering analysis; optimal selection, the average entropy ratio is calculated, and an index with the highest significance is reserved and selected in combination with a clustering analysis map. According to the method, the indexes are selected in an open manner, the design is reasonable, and more scientific and more effective battery consistency evaluation indexes are obtained through a series of quantitative analysis.

A system for storing a specification document to be checked on a storage device in a computer readable format. A computer implemented method which stores a graph structure of rules for performance of checking, and creates a rule application table of whether a rule is an application subject. For each node rule of the graph structure, average entropy of whether a specification satisfies the rules is derived by searching the graph structure of rules. While performing a depth-first searching of the graph structure of rules, a priority order index is derived from the rule application table for each node rule. Amendment candidates are then displayed according to value of the priority order index.

The invention belongs to the technical field of information security and true random number generators, and particularly relates to a test method for a maximum entropy rate of a true random number generator. The test method for the maximum entropy rate of the true random number generator comprises the four steps of parameter selection, data collection, block segmentation calculation and analysis fitting. By utilizing a characteristic that an average entropy rate obtained by calculation after block segmentation of a random data stream is saturated along with the reduction of the data block length and the increment of a data rate, a relational curve of a saturation entropy rate and a data block length is subjected to linear fitting; a function value corresponding to a fitting straight line when the data block length is equal to 1 is namely the maximum entropy rate; and the problem that the maximum entropy rate in a test of the true random number generator cannot be uniquely determined can be effectively solved.

The invention relates to a stock market investor sentiment index selection method based on a clustering and average entropy difference group, which relates to the field of stock market investor sentiment measurement, comprising the following steps: index selection: any index related to stock market or investor can be brought into a candidate index database; primary selection, invalid indicators whose data are not available or whose data are not continuous are deleted; screening: irrelevant indexes are rejected by irrelevant analysis; selection: through correlation analysis and clustering analysis, redundant information more than the threshold of the index is removed. Preferably, the difference between the information entropy and the average entropy in various states is calculated to obtainthe clusters, and the index with the highest significance is retained by combining the cluster analysis map. The open selection index of the invention is reasonable in design, and through a series ofquantitative analysis, a more scientific, more inclusive and more extensive proxy index of investor sentiment measurement is obtained.

The invention relates to a stock market investor sentiment index selection method based on the ratio of clustering and maximum entropy increment, which relates to the field of stock market investor sentiment measurement. The method comprises the following steps: index selection; any index related to stock market or investor can be brought into a candidate index database; primary selection, excluding invalid indicators whose data are not available or whose data are not continuous; screening, rejecting irrelevant indexes by irrelevant analysis; selection, through correlation analysis and clustering analysis to remove redundant information more than the threshold of the index; preferably, the sensitivity factor, i.e. the ratio of information entropy ratio average entropy to maximum entropy ratio average entropy, is calculated, and the index with the highest significance is reserved by combining the clustering analysis map. The open selection index of the invention is reasonable in design,and through a series of quantitative analysis, a more scientific, more inclusive and more extensive proxy index of investor sentiment measurement is obtained.

The invention discloses a cognitive attribute selection method and system, a category attribute determination method and system and a storage medium. The cognitive attribute selection method comprises the following steps: obtaining values of attributes in a sample set; calculating the probability of taking each value corresponding to each attribute in the sample set; calculating extra information amount obtained when each attribute in the sample set takes each value corresponding to the attribute; determining the average entropy subtraction of each attribute in the sample set according to the probability and the additional information amount; and taking the attribute corresponding to the maximum average entropy subtraction in the average entropy subtraction as the selected attribute. The attribute corresponding to the maximum average entropy subtraction is taken as the preferable cognitive attribute, and the most information about the category attribute of the cognitive target can be obtained through the value screening sample set of the preferable cognitive attribute of the cognitive target, so that the category attribute of the cognitive target can be determined as soon as possible, and the cognitive efficiency of the category attribute of the cognitive target is improved.

A method, computer program product, and computing system for processing a data portion to divide the data portion into a plurality of data chunks; performing an entropy analysis on each of the plurality of data chunks to generate a plurality of data chunk entropies; and determining an average data chunk entropy from the plurality of data chunk entropies.

The invention discloses an automatic white balance processing method and device for a large scene range. The method comprises the following steps: acquiring image data; dividing the image data into aplurality of equal modules; calculating the average entropy of the image data and the entropy of each block of the plurality of equally-divided modules; calculating a weighted gray value of the imagedata according to the average entropy and the entropy of each block; and performing white balance processing on the image data according to the weighted gray value. A color temperature sensor does notneed to be added or a filter does not need to be added in the camera, use of software resources can be reduced while hardware cost is reduced, and cost is saved. Compared with a traditional correction mode, the method used by the invention still has a relatively good effect in a scene with a single color, and the image color can be correctly restored when the illumination is low or the exposure is abnormal.

The present invention proposes an entropy-based neural machine translation dynamic decoding method and system. By analyzing the relationship between the entropy value of a sentence and the BLEU value, it is found that the average entropy value of words in a sentence with a high BLEU value is lower than that in a sentence with a lower BLEU value. The average entropy value of words is small, and the BLEU value of sentences with low entropy value is generally higher than that of sentences with high entropy value. By calculating the Pearson coefficient between the entropy value of the sentence and the BLEU value, it is found that there is a correlation between the two. Therefore, the present invention proposes that in each time step of the decoding phase of the training process, not only must a certain probability be sampled to select real words or predicted words to obtain context information, but also to calculate the entropy value according to the prediction result of the previous time step, and then according to the entropy The value dynamically adjusts the weight of contextual information. Addresses the issue of error accumulation in neural machine translation models during decoding due to differences in contextual information between training and inference.

The invention discloses a quality assessment method based on trajectory data, which includes steps: S1, expressing the trajectory data as a set of tuples in four dimensions, and obtaining a map; S2, grouping the trajectory data according to entities, and using the time dimension to The grouping is sorted, and each trajectory of the trajectory data is converted into a set of ordered line segments; S3, dividing the map into grids of equal size, and projecting the trajectory data into the grid; S4, calculating entropy of the grid, and calculate the average entropy and weighted average entropy of all the grids. The invention utilizes entropy to effectively evaluate the quality of trajectory data, so as to realize the comparison of the data quality of multiple pieces of trajectory data and analyze the feasibility of using it for data analysis, and the data quality assessment is simple, effective and low-cost.

The invention discloses a recommendation system attack detection method based on time series data. The method includes the steps that a user-project preference degree data set and a frequent item set excavation technology are utilized to acquire user groups and project groups; group preference degree value proportional characteristics are calculated for the pairs of user sets and the pairs of project sets; all preference degrees of projects in the project groups form time series preference degree data according to operating time; group preference degree time interval characteristics are calculated for the pairs of user sets and the project sets; group average entropy characteristics are calculated for the user groups; for the user groups, maximum group preference degree value proportional characteristics corresponding to the user groups and maximum group preference degree time interval characteristics corresponding to the user groups are selected, and the user groups are sequentially ranked through the three characteristics, and then three ordered user group sequences are acquired; the three ordered user group sequences are integrated to acquire a wholly ordered user group sequence, and the most probably attack user groups are acquired; the most probably target project groups are acquired through the group preference degree value proportional characteristics.

The invention relates to a maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method and belongs to the image matching field. Scale-invariant?feature?transform (SIFT) operators have strong matching ability, while, the scale-invariant?feature?transform (SIFT) operators will bring a huge amount of data, and therefore, binaryzation should be performed on the scale-invariant?feature?transform (SIFT) operators, however, if unified binaryzation is performed on all the operators, data redundancy or information loss will be brought about. The maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method of the invention comprises the following steps that: binaryzation is performed on the scale-invariant?feature?transform (SIFT) operators; average entropy calculation is performed on each layer of binaryzation results, such that different numbers of binaryzation layers are selected adaptively; new binaryzation descriptors are provided; the Hamming distance is utilized to replace the Euclidean distance so as to calculate the distance between two descriptors; and the distance between the two descriptors is compared with a set threshold value. With the maximum average entropy-based scale-invariant?feature?transform (SIFT) descriptor binaryzation and similarity matching method of the invention adopted, information of original features is reserved; the amount of data storage can be greatly reduced; computational complexity can be reduced; a requirement for a real-time property can be realized better; the matching results equivalent to original scale-invariant?feature?transform (SIFT) descriptors can be obtained and are far superior to results of matching by using the unified binaryzation.

The invention discloses a fast detecting method and device for a static vehicle target. The fast detecting method includes the following steps that images are collected, wherein vehicle images are obtained; the images are preprocessed, wherein normalization is conducted on the sizes of the vehicle images, and the vehicle images are converted into a grey-scale map; entropy information is extracted, wherein pixel level traversal is conducted on the images through search boxes, and pulse iteration average entropy information of the partial images in the search boxes is extracted; rough positioning is conducted on the vehicle target, wherein the images with the pulse iteration average entropy value larger than that the target entropy value are taken as vehicle target rough positioning images; the vehicle target is determined, wherein the vehicle image with the largest pulse iteration average entropy value in the same picture area is reserved as a target vehicle search result; the vehicle target is presented in an original large image, and positioning of the vehicle target is completed. According to the fast detecting method and device for the static vehicle target, robustness, rapidness and transportability of a pulse coupling neural network feature extraction method are used, the feature of the pulse iteration average entropy is selected as the judgment basis, and the speed of detecting the static vehicle target is greatly increased.

The invention relates to a stock market investor sentiment index selection method based on principal components and average entropy increments, which relates to the field of stock market investor sentiment measurement. The method comprises the following steps: index selection: any index related to stock market or investor can be brought into a candidate index database; primary selection: invalid indicators whose data are not available or whose data are not continuous are excluded; screening: irrelevant indexes are excluded by irrelevant analysis; selection: through correlation analysis and clustering analysis, redundant information more than the threshold of the index is removed; preferable selection: the principal component analysis is used to calculate the sensitivity factor, that is, the difference between the information entropy and the average entropy, the average entropy, and the cluster analysis is used to select the index with the highest significance. The open selection indexis reasonable in design, and the reserved quantity of the index is determined through a series of quantitative analysis, so a more scientific, more inclusive and broader proxy index for investor sentiment measurement is obtained.

The invention discloses a super-pixel-based quaternion wavelet domain image watermark detection method. Entropy rate super pixel segmentation is carried out on a host image; smooth super pixels and non-smooth super pixels are divided based on average entropy values; a noise point suppression threshold of an SIFER operator is determined adaptively by combining the average entropy value of the superpixels; image feature points are extracted in texture super pixels and the smooth super pixels respectively by using the SIFER operator; an elliptic feature region is constructed by using an auto-correlation matrix R and a final stable elliptic region is screened out by combining an attack result; the constructed elliptic feature region is mapped to be a circular feature region, 0 is supplementedaround the circular feature region to obtain an externally connected square sub image, first-order quaternion wavelet transform is carried out on the square region, and an amplitude matrix and a phase matrix are calculated; and then with a quantization method, watermark embedding is carried out by using the amplitude as a carrier.

The invention relates to a burst traffic detection method based on dynamic threshold. By means of the method, the burst traffic in a high-speed network environment can be effectively detected on line in real time, and the problem that a fixed threshold method cannot well adapt to the dynamically changing network environment is solved. According to the technical scheme, a threshold is dynamically adjusted according to the source IP entropy value of the normal historical traffic, a sliding window mechanism is introduced, and burst traffic is detected by comparing whether the difference value between the current detection value and the average entropy value of a sliding window exceeds the dynamic threshold; the method comprises steps as follows: generation of an initial sliding window, generation of a reference value, generation of the current detection value, abnormality judgment of the current detection value, processing of normal traffic and processing of burst traffic. The detection method is rapid, simple, high in detection rate and low in false rate, the network safety is effectively guaranteed, and the detection method has high practicability and huge economic and social benefits.