48 results about "Forward selection" patented technology

The invention takes characteristics of a wireless mesh network as a clue, combines a forwarding strategy of opportunistic routing, and provides a routing method in application programming (IAP) which is applicable to the wireless mesh network by improving a greedy perimeter stateless routing (GPSR) routing protocol. The IAP is functionally divided into two parts: 1) in a greedy mode, a data acquisition module and a data acquisition interface are arranged on an IEEE 802.11 multi-access computer (MAC) layer, the state information of a local node interface and link information between neighboring nodes are acquired on the MAC layer, the processed data is fed back to a routing layer in a call-back mode, and finally, comprehensive routing metrics are formed, so that packet node forwarding selection in the greedy mode is closer to actual conditions of the network; and 2) when the packet meets a routing hole in the wireless mesh network, a novel competition-based opportunistic routing hole detouring strategy is provided; moreover, on a hole edge, the neighboring nodes compete forwarding opportunity according to the mastered topology and link information to reduce the possibilities of increasing routing hops brought by blind detouring and generating network congestion.

The invention provides an automatic sleep stage classification method based on dual character filtering. The stage classification method comprises steps of firstly extracting a two-guidance sleep electroencephalogram signal and an one-guidance horizontal electrooculogram signal; performing filtering for the original electroencephalogram signals and electrooculogram signal; extracting multiple characters from the filtered electroencephalogram signals and the filtered electrooculogram signal; selecting the optimal character subsets by use of a dual character filtering method combining the Fisher score method and the sequential forward selection method. Via the dual characteristic filtering method, character dimension is greatly reduced and redundancy among the characters is reduced. At last, a support vector machine classifier is used for identifying the optimal characters, so automatic stage classification of sleep is finished. According to the invention, objectivity, precision and convenience of automatic sleep stage classification can be well increased; the automatic sleep stage classification method is characterized by high precision, low calculation complexity, simple operation and easy popularization; and considerable social and economic benefit can be gained.

The invention relates to the technical field of biological information, and discloses a method and system for predicting amino acid mutation. The method and system for predicting amino acid mutation aim at improving the accuracy and the effect of prediction and effectively solving the problems that bioexperiment is blind, the cost of the bioexperiment is high and the like. The method for predicting amino acid mutation comprises the steps of establishing a protein sample set; determining characteristics of prefiltering, and integrating characteristics of the same sample into one characteristic sequence to combinedly establish an initial characteristic set of the sample; screening out relatively important characters through a stable character selection algorithm to combinedly establish a first screening out characteristic set of the sample; screening out important characters through a sequence forward selection algorithm to combinedly establish the final screening out characteristic set of the sample; selecting a positive sample and a negative sample to establish a training set and an independent test set, substituting the final screening out characteristic set of samples in the training set into a gradient promoting tree algorithm to be subjected to training so as to obtain a final disaggregated model, and conducting assessment on a prediction result of the disaggregated model by combining the final screening out characteristic set of the independent test set.

The invention provides a visual-auditory evoked emotion recognition method and a system based on an EEG signal. The method comprises the following steps: firstly, the EEG signal generated based on thevisual-auditory evoked emotion is collected and preprocessed through a band-pass filter. Then the multichannel EEG signals are decomposed by multivariate empirical mode decomposition, and the effective intrinsic mode functions are selected to extract the features of emotional EEG signals. Then, the sequence floating forward selection method is used as the search strategy to select and delete feature sets, and the filter and encapsulator are used as the evaluation criteria of the optimal feature subsets to filter the extracted emotional EEG features. Then the selected feature subset is input to support vector machine for classification, and the classification results are obtained. Finally, according to the classification results, the emotion recognition results are obtained, and the emotion recognition is realized. On the basis of exploring the law of emotional electroencephalogram, the invention carries out the research on the emotion recognition method of a plurality of induced modes, and effectively improves the recognition accuracy.

The invention discloses a hyperspectral characteristic variable selection method and provides a characteristic forward selection support vector machine (SVM-BFFS) which is an embedded type forward sequence selection method base on the SVM maximum interval principle. The hyperspectral characteristic variable selection method is popularized to support the SVM-BFFS of a variety of classification problems by utilization of a one-to-one strategy. According to the hyperspectral characteristic variable selection method, stability of characteristics is ranked by exploring the internal relation between characteristic variables and the stability, and therefore the characteristic variables of infrared hyperspectra can be selected quickly and effectively. By application of the hyperspectral characteristic variable selection method, dimensionality of a hyperspectral image can be reduced, and therefore complexity of follow-up calculation is reduced, and a high accuracy rate of a built model is maintained.

The invention discloses a polarization feature selection and classification method based on an object-oriented random forest, which solves the problem of feature selection and image classification when a plurality of multi-polarization features participate in classification. The method adopts the object-oriented method to segment a feature set in multi-scale, performs random forest modeling on thesegmented sample objects, calculates the importance of each feature, and optimizes the feature set by using the forward selection algorithm. The invention adopts the object-oriented random forest method to improve the model training efficiency and the classification accuracy. The sequence forward selection algorithm and the highest precision iterative termination condition are used to construct the optimal feature subset to avoid falling into the local optimal solution. The algorithm can improve the classification accuracy, and provide quantitative reference for the reasonable optimization offeature set.

The invention belongs to the technical field of image processing, and specifically relates to a lightweight incisional hernia patch three-dimensional ultrasonic image feature extraction method. The method comprises the steps of firstly extracting related textural feature parameters of regions to be classified in three-dimensional volume of interest (VOI) of an automated three-dimensional breast ultrasound (ABUS) image in automatic quantification manner by using textural feature extraction algorithm so as to be used for differentiating a patch and a fascia; then introducing three-dimensional textural parameters and three-dimensional location parameters to improve the robustness of a lightweight patch classification and recognition algorithm in allusion to a problem that two-dimensional textural parameters are sensitive to spatial transformation such as post-operation curling and contraction of an incisional hernia patch; and finally performing feature selection by using a distance-between-class algorithm and a sequential forward selection method. The method provided by the invention is good in feature selection effect, high in efficiency, capable of effectively improving the classification accuracy of lightweight incisional hernia patch three-dimensional ultrasonic images, and convenient for automatic classification and recognition.

The invention discloses a driver fatigue driving monitoring and warning method and system based on the Internet of Vehicles. The method comprises steps that driving parameters are simultaneously collected; the driving parameters are pre-processed, feature extraction of the driving parameters is performed, and a full feature set is determined; features of the full feature set are normalized, the features in the full feature set are filtered through the sequence floating forward selection algorithm, and the optimal feature subset is determined; the optimal feature subset is taken as the input ofa support vector machine, a driving state of a driver is taken as the output of the support vector machine, a fatigue detection model is constructed; a current driving state of the driver is determined according to the fatigue detection model; the current driving state is transmitted to a client through utilizing the Internet of Vehicles technology in real time, and monitoring is performed by a user of the client or an enterprise. The method is advantaged in that the false identification rate of the monitoring system can be reduced, and a problem of influence on the driving comfort due to wearing a detection device is solved.

Systems and methods provide for feature selection that combines semantic classification and generative filtering with forward selection. Features from an original feature set are divided into feature subsets corresponding to ranked semantic classes. Additionally, low quality features are removed from consideration. Features are selected for a reduced feature set by iteratively processing the feature subsets using forward selection in an order corresponding to the ranking of the semantic classes. The reduced feature set is used to generate a predictive model.

The invention discloses a real-time communication routing method for a wireless sensor network, and the real-time communication routing method comprises the steps of obtaining the transmission speed of data packets between a current node and all predecessor neighbor nodes of the current node by calculation according to communication delay between the current node and the predecessor neighbor nodes of the current node and queuing delay of the predecessor neighbor nodes after the current node receives the data packets which are sent from an upstream node to a base station node, taking the obtained transmission speed of the data packets as forwarding selection standard, and selecting the next hop of the node in a predecessor forwarding neighbor table to forward the data packets. The real-time communication routing method can effectively ensure that the data packets can select a path with higher efficiency for forwarding during the real-time forwarding process, reduce transmission time delay of the data packets and simultaneously improve the transmission reliability of the data packets.

The invention discloses a greenhouse gas emission monitoring method of a town household refuse disposal system. By monitoring methane gas and refuse components and revising the refuse components, degradable organic carbon content, proportion of methane in refuse landfill gases, a half-life period and the like, the greenhouse gas emission monitoring method puts forward selection principles of the parameters under different conditions selectively, performs connection on instruments on site and upper computer working stations and is capable of updating data in real time. The greenhouse gas emission monitoring method reduces monitoring cost, is stable in monitoring results and high in accounting accuracy.

The invention provides a model transfer method based on hyperspectral data. The method comprises the steps of using a primary instrument and a secondary instrument to perform spectral collection on a sample, and establishing a classified prediction model of the sample according to spectral data collected by the primary instrument after all the spectral data is subjected to preprocessing; selecting N groups of corresponding spectral data as a conversion data set, and building a spectral transfer relationship; conducting correction on spectral data collected by the secondary instrument according to the spectral transfer relationship; randomly selecting M spectral data from corrected spectral data of the secondary instrument as a training sample, and regarding the rest data as a testing sample; picking m dimensional data with highest accuracy from n dimensional spectral data to conduct waveband combination according to a forward selection algorithm; classifying the m dimensional data according to a machine learning method to determine the classification of the testing sample and calculating the accuracy of the testing data. By means of the model transfer method based on the hyperspectral data, high-accuracy classified prediction of the spectral data collected by different instruments can be achieved, and meanwhile the processing time of a data classification algorithm is shortened.

A tuner receives broadcast data on specified channels. A channel search unit controls the tuner to search for available channels, and creates a channel map representing the available channels and a selection order thereof. A channel order setting unit changes setting about enabling / disabling of the selection and / or the selection order of the channels in the channel map. An operation input unit accepts a channel selecting operation of a user. When the user performs an operation for channel forward selection or channel reverse selection through the operation input unit, a channel switching unit changes the selection of the channel according to the channel map and causes the tuner to receive the broadcast data on the selected channel.

The invention relates to systems and methods for prioritizing video slices of H.264 video bitstream comprising: a memory storage and a processing unit coupled to the memory storage, wherein the processing unit operates to execute a low complexity scheme to predict the expected cumulative mean squared error (CMSE) contributed by the loss of a slice of H.264 video bitstream, wherein the processing unit operates to execute a series of actions comprising assigning each slice a predicted value according to the low complexity scheme; extracting video parameters during encoding process, said video parameters; and using a generalized linear model to model CMSE as a linear combination of the video parameters, wherein the video parameters are derived from analytical estimations by using a Generalized Linear Model (GLM) over a video database, encompassing videos of different characteristics such as high and low motion, camera panning, zooming and still videos, further comprising wherein the GLM is constructed in a training phase as follows: determining the distribution of the computed CMSE to be a Normal distribution with the Identity link function; sequentially adding covariates using the forward selection technique where by the best model is evaluated at each stage using the Akaike's Information Criterion (AIC); the training phase of the model generates regression coefficients; the final model is validated through the testing phase by predicting the CMSE for different video sequences, not in the training database; and by using the regression coefficients, the expected CMSE values are predicted for each slice.

The invention discloses a fatigue driving detection-oriented steering wheel operation feature extraction method. The method includes the following steps that: statistical analysis and comparison are performed on driver fatigue sample variables; the significance of difference of fatigue discrimination indexes under different fatigue levels is tested according to analysis results, and significance indexes are selected to construct fatigue discrimination indexes; with the classification performance of a support vector machine algorithm adopted as evaluation criteria, and a sequential floating forward selection algorithm adopted as a search strategy, a fatigue discrimination index approximate optimal selection algorithm is established, and an index system of driver fatigue state detection is established; with the index system of the driver fatigue state detection obtained through screening adopted as input, a driver fatigue state detection model can be built based on the support vector machine algorithm; and a driver fatigue state detection model considering individual difference and a fatigue detection model in lane deviation of vehicles are built based on the support vector machine algorithm. The method is suitable for generalization ability of different drivers and different operation states and can improve the recognition accuracy of the fatigue detection model.

The invention discloses an editing operation system and method of pure text content. The method provided by the invention comprises the following steps of: selecting a selection range of the pure text content by a touch display device according to a forward selection sequence and selecting contents to be deleted in the selection range according to the forward selection sequence; repeatedly carrying out movements of a secondary slippage in the touch display device according to a backward selection sequence to recover the content to be deleted which is deleted at the last time. Thereby, the technical efficacy for rapidly and conveniently carrying out the editing operation on the pure text content is supplied to a handheld type device.

The invention discloses a boiler NOx prediction method based on MI-LSTM. The method comprises: ranking the importance of auxiliary variables and selecting variables by using mutual information (MI) "minimum redundancy maximum correlation" as a criterion; using a selected auxiliary variable set as an input of a long short term memory (LSTM) model; establishing an MI-LSTM model for NOx emission prediction, and searching for an optimal parameter for the LSTM model by using multilayer grids. In the process of variable selection, a sequence forward selection method is used to determine the optimalinput feature set and the optimal model parameter with the model prediction accuracy as a goal, thereby effectively reducing the information redundancy between the input variables, reducing the modelcomplexity, and improving the prediction accuracy and generalization capability of the model.

The invention discloses a film evaluation prediction method and system. The method comprises the steps of firstly selecting characteristics of film data set for training through a sequential forward selection method, and conducting de-noising, pre-processing and extracting of a characteristic vector; secondly collecting rating information corresponding to each film, and conducting normalization processing of all the ratings to obtain a rating distribution vector; thirdly, training a maximum interval rating distribution model based on the film characteristic vector and the rating distribution vector, and conducting deformation optimization to obtain a final parameter model for film rating prediction; and fourthly, extracting a characteristic vector of a film to be released, obtaining a vector by using the obtained parameter model for rating distribution prediction, conducting normalization processing of the vector, and obtaining the predicted result of the rating distribution of audience after the film is released. According the invention, the model for film rating distribution prediction can be quickly and effectively trained, and the accuracy of prediction is substantially higher than that by using existing methods.

The invention relates to an industrial process fault diagnosis method based on the Bayesian information criterion. The method comprises: collecting normal industrial data and calculating several kindsof detection statistics amounts based on normal data; carrying out fault detection on a to-be-detected sample; expressing a fault isolating task into a combinatorial optimization problem; convertingthe problem into a mixed integer nonlinear programming problem by combining the Bayesian information criterion; on the basis of a forward selection algorithm, simplifying the problem into a mixed integer quadratic programming problem; on the basis of a branch-and-bound algorithm, solving a series of similar mixed integer quadratic programming problem to obtain a fault variable combination causingthe sample fault. The industrial process fault diagnosis method has high universality; and the fault variable can be identified without predetermining a fault direction or a known historical fault data set. When the amplitude of the fault is small, an accurate diagnosis result is obtained. Besides, the combination optimization problem is transformed into the quadratic programming problem with sparse constraints for calculation, so that the computational efficiency is improved substantially.

The invention discloses a user identity recognition method based on a carrying position and a carrying mode of a smart mobile phone, and belongs to the technical field of mobile terminal control recognition. The method comprises the following steps: firstly, acquiring sensor data of the mobile phone at different positions; filtering the sensor data by a SavitzkyGolay filter; partitioning an acceleration time sequence into data segments by using a gait cycle detection algorithm; extracting time domain features and frequency domain features from the partitioned data by adopting a statistics method; extracting a feature set of position recognition by using a SFFS (Sequential Floating Forward Selection) method; confirming the specific position of the mobile phone worn by a user according to the selected features in conjunction with a DT algorithm; and finally, converting codes of the recognized mobile phone wearing position through a One Hot Coding algorithm, and performing user identity recognition according to extracted statistic features in conjunction with trained user information by a SVM (Support Vector Machine) algorithm. Through adoption of the user identity recognition methodprovided by the invention, a continuous, safe and reliable identity recognition method is provided for mobile phone users and application manufacturers.

A method of feature selection applicable to both forward selection and backward elimination of features is provided. The method selects features to be used as an input for a classifier based on an estimate of the area under the ROC curve of each of the classifiers. Exemplary applications are in homecare or patient monitoring, body sensor networks, environmental monitoring, image processing and questionnaire design.

The invention provides an adaptive alternating successive approximation analog-to-digital converter with low power consumption, comprising a comparator, capacitor arrays and a control logic circuit, wherein a positive input end of the comparator is connected with a forward capacitor array, and a negative input end of the comparator is connected with a reverse capacitor array; an output end of thecomparator is connected with the control logic circuit; an upper polar plate of the forward capacitor array is connected with a positive input voltage, and a lower polar plate of the forward capacitorarray is respectively connected to a reference voltage and is grounded through a forward selection switch; the upper polar plate of the reverse capacitor array is connected with a negative input voltage, and the lower polar plate of the reverse capacitor array is respectively connected to the reference voltage and is grounded through a reverse selection switch; and the forward selection switch and the reverse selection switch are both connected with the control logic circuit. The invention also provides a corresponding control method. The adaptive alternating successive approximation analog-to-digital converter with low power consumption provided by the invention has the advantages that the design difficulty of the digital logic circuit can be reduced and the overall power consumption canbe reduced.

The invention provides a method of probabilistic load flow calculation for a power system based on scene reduction. The method includes the steps of performing Monte Carlo sampling on random factors such as fans, photovoltaic energy, energy storage devices and controllable loads in a power system to obtain the operating state of the power system and to construct an initial scene library; reducing the initial scene library using a fast forward selection method to filter out typical scenes, completing the classification of the scenes and generating scene sets; on this basis, carrying out probabilistic load flow calculation for each scene set using a semi-invariant method to obtain probability distribution characteristics of the state quantity of a power grid; and overlapping the load flow characteristics of each scene set to obtain the overall load flow distribution characteristics of the power grid. The method provided in the invention can be applied to the load flow calculation of a large-scale complex power grid with a high proportion of renewable energy and can effectively solve the problem of a precision decrease in the load flow calculation result by the semi-invariant method due to the large fluctuation of the random factors in the power system, thereby ensuring the high accuracy of the probabilistic load flow calculation of the power grid.

A tuner receives broadcast data on specified channels. A channel search unit controls the tuner to search for available channels, and creates a channel map representing the available channels and a selection order thereof. A channel order setting unit changes setting about enabling / disabling of the selection and / or the selection order of the channels in the channel map. An operation input unit accepts a channel selecting operation of a user. When the user performs an operation for channel forward selection or channel reverse selection through the operation input unit, a channel switching unit changes the selection of the channel according to the channel map and causes the tuner to receive the broadcast data on the selected channel.

The invention discloses a real-time communication routing method for a wireless sensor network, and the real-time communication routing method comprises the steps of obtaining the transmission speed of data packets between a current node and all predecessor neighbor nodes of the current node by calculation according to communication delay between the current node and the predecessor neighbor nodes of the current node and queuing delay of the predecessor neighbor nodes after the current node receives the data packets which are sent from an upstream node to a base station node, taking the obtained transmission speed of the data packets as forwarding selection standard, and selecting the next hop of the node in a predecessor forwarding neighbor table to forward the data packets. The real-time communication routing method can effectively ensure that the data packets can select a path with higher efficiency for forwarding during the real-time forwarding process, reduce transmission time delay of the data packets and simultaneously improve the transmission reliability of the data packets.

The invention discloses a sensor network data distribution forward selection method based on network coding. The average effective reception amount of all coded data packets to be sent is calculated by a forward node, and according to the average effective reception amount, the sending of the requested coded data packets is started after a waiting time, wherein the average effective reception amount is larger, the waiting time is shorter. According to the method of the invention, compared with the prior art, the link correlation influences that are in objective existence are taken into account, so forward selection can be performed more accurately.

A control device for power supply time of timers comprises an alternative current (AC) power source input module, a direct current (DC) power source supply module, an AC power source output module electrically connected to a load end, a relay, a central control module, a relay driver circuit, a day / night detecting module, a power supply time backward display light set, a power supply time forward display light set, a power supply time backward selection control button and a power supply time forward selection control button. Herein, by pressing the power supply time backward selection control button or the power supply time forward selection control button, one of the power supply time backward display lights or the power supply time forward display lights can be turned on such that the central control module can, based on the sunset time of the previous day as well as the configured power supply backward setting time or power supply forward setting time, postpone or advance the power supply time for the load end in the next day, and then record the sunset time of the next day so as to control the power supply time for the load end in the further next day.

The invention relates to a feature selection method based on filtering type and packaging type hierarchical progression. The method comprises the following steps: firstly, sorting features by using a filter type variance sorting method, an information gain sorting method and a Boruta sorting method based on a packaging type; distributing ranks to the sorted features according to importance degrees; fusing results of the three sorting methods; and then calculating correlation between every two features based on a Pearson's correlation coefficient, so as to obtain a feature fusion result; then setting a Pearson's correlation coefficient threshold of the features, selectively deleting part of the features according to the correlation between the features, and finally finding out the best feature combination based on a packaged sequence forward selection method in combination with a random forest model, thereby obtaining an optimal feature subset. The method has a good effect of selecting the optimal feature subset for the data set, and provides relatively accurate feature information for the learning model, so that the accuracy of the learning model is improved.

A novel metric, termed cluster resolution, which compares the separation of clusters of data points while simultaneously considering the shapes of the clusters and their relative orientations. This metric, in conjunction with an objective variable ranking metric, allows for the fully-automated determination of the optimal number of variables to be included in a chemometric model of a system. Cluster resolution is based upon considering the minimum distance between (or the extent of overlap of) confidence ellipses constructed around clusters of points representing different classes of objects. In one implementation, PCA was used to classify samples of gasoline based upon their octane rating. The entire GC-MS chromatogram of each sample comprising over 2×106 variables was considered. These variables were ranked by ANOVA and then a forward selection approach was used to find the optimal number of variables for inclusion. In this implementation, a model having 2761 variables was found to be optimal. This approach can be generally applied to feature selection for a variety of applications and represents a significant step towards the development of fully-automated, objective construction of chemometric models.