57 results about "One-class classification" patented technology

A classification apparatus, method, and computer program product for multi-modality classification are disclosed. For each of a plurality of modalities, the method includes extracting features from objects in a set of objects. The objects include electronic mail messages. A representation of each object for that modality is generated, based on its extracted features. At least one of the plurality of modalities is a social network modality in which social network features are extracted from a social network implicit in the set of electronic mail messages. A classifier system is trained based on class labels of a subset of the set of objects and on the representations generated for each of the modalities. With the trained classifier system, labels are predicted for unlabeled objects in the set of objects.

The present invention provides a method for recognizing a named entity included in natural language, comprising the steps of: performing gradual parsing model training with the natural language to obtain a classification model; performing gradual parsing and recognition according to the obtained classification model to obtain information on positions and types of candidate named entities; performing a refusal recognition process for the candidate named entities; and generating a candidate named entity lattice from the refusal-recognition-processed candidate named entities, and searching for a optimal path. The present invention uses a one-class classifier to score or evaluate these results to obtain the most reliable beginning and end borders of the named entities on the basis of the forward and backward parsing and recognizing results obtained only by using the local features.

Disclosed herein is an anomaly detection method for a packet-based network which includes several network resources, also called network-related software objects. The method includes monitoring the network resources of the packet-based network, ordering the monitored network resources according to a given ordering criterion, and detecting an anomaly in the packet-based network based on the ordered network resources. In particular, detecting an anomaly includes forming a detection feature vector based on the ordered network resources, and feeding the detection feature vector to a machine learning system configured to detect an anomaly in the packet-based network based on the detection feature vector. The detection feature vector includes detection feature items related to corresponding monitored network resources, and arranged in the detection feature vector depending on the ordering of the corresponding monitored network resources. Conveniently, the machine learning system is a one-class classifier, preferably a one-class Support Vector Machine (OC-SVM).

Aspects of the present invention relate to a method of sharing content between communication devices in a communication network. The content sharing may be allowed and / or managed by a network administration module in the communication network. In this regard, the method may comprise the following steps: (1) receiving a first type of classification information related to a content emanating from a first communication device, the first type of classification information classifying the content; (2) determining where the classified content emanating from the first communication device is located; (3) determining whether the classification of the content corresponds to a second type of classification linked with a second communication device; and (4) in the case where the classification of the content corresponds to the second type of classification, then making the content available to the second communication device.

The invention belongs to the field of automatic detection and control of textile quality and particularly relates to a method for automatically detecting defects of textiles based on support vector data description theory. The automatic textile defect detection based on computer vision is a more difficult one-class classification task in real world. In the invention, support vector data description (SVDD) of the advanced one-class classification method is applied in the textile defect detection field for the first time to obviate the problems present in textile defect detection of conventional two-class classification support vector machine, which is difficult to collect representative defect samples completely and at larger number and further fails to effectively train the detector. Additionally, the invention provides a robust new method for solving the optimization problem of parameters, in particular to scale parameter of gauss kernel function, related in SVDD training. The automatic textile defect detection system based on SVDD can prospectively and conveniently control false alarm rate (false detection rate) in practice and can obtain lower miss ratio at lower false alarm rate.

An apparatus for monitoring of a device including a moveable part, especially a rotating device, wherein the apparatus includes a control module which receives a measured vibration signal of the device provided by a sensor connected to the device, provides a spectrum of the measured vibration signal, pre-processes the spectrum to determine base frequencies and side frequencies, where the base frequencies are frequencies having peak powers corresponding to eigen frequencies of the device or faulty frequencies and the side frequencies correspond to other frequencies, where the control module additionally processes the base and side frequencies by applying separately a one-class classification on the base and side frequencies, combines the results of the one-class classifications to obtain a classification signal representing a confidence level, and outputs a decision support signal based on the classification signal, where the decision support signal indicates an error status of the monitored device.

The invention provides a face detection system for multi-pose faces. The face detection system comprises a front end detector and a back end detector. The front end detector comprises at least one layer of classifier. Each layer comprises at least two parallel first type classifiers for different poses of faces, wherein the first type classifiers are used for distinguishing candidate faces and non-face windows. The back end detector comprises a second type classifier using a depth nerve network. The second type classifier is used for further distinguishing faces and non-faces in detection results of the front end detector. Accordingly, the invention also provides a face detection method. While the detection precision is improved, calculation expense in the detection process is effectively reduced and detection speed is effectively increased.

The invention discloses a sensitive-image identification method and a system, and belongs to the technical field of image identification. The sensitive-image identification method and the system are characterized in that the following steps are comprised: a step 1, grid dividing characteristic extraction fused with skin color detection is carried out, and original bag-of-words expressing vectors of images are obtained through a bag-of-words model; a step 2, image characteristic optimization is carried out, and dimension-reduced optimization image vector expressions are obtained through the utilization of a random forest; a step 3, identification model training is carried out, that is to say through the utilization of a one-class support vector machine, a one class classifier is trained in optimization vector space; and a step 4, image identification is carried out, i.e., if the images completely do not contain skin color pixels in the pretreatment process of the step 1, the images are directly determined to be normal images; and otherwise, optimization characteristic expressions are obtained after processing, and the optimization characteristic expressions enter the one-class classification model obtained through the training, so that identification results of the images are finally obtained. According to the invention, a one-class classification algorithm is utilized to solve sensitive-image identification problem, and a plurality of techniques are fused in the processing process, and the characteristic optimization processing is carried out, so that the accuracy and the efficiency of the sensitive-image identification are improved.

The invention relates to an incremental learning-fused support vector machine multi-class classification method, and aims to reduce sample training time and improve classification precision and anti-interference performance of a classifier. The technical scheme comprises the following steps of: 1, extracting partial samples from total samples at random to serve as a training sample set D, and using the other part of samples as a testing sample set T; 2, pre-extracting support vectors from the training sample set D; 3, performing support vector machine training on a pre-extracted training sample set PTS by using a cyclic iterative method so as to obtain a multi-class classification model M-SVM; 4, performing binary tree processing on the multi-class classification model M-SVM to obtain a support vector machine multi-class classification model BTMSVM0; 5, performing incremental learning training on the multi-class classification model BTMSVM0 to obtain a model BTMSVM1; and 6, inputting the testing sample set T in the step 1 into the multi-class classification model BTMSVM1 for classification. The incremental learning-fused support vector machine multi-class classification method is used for performing high-efficiency multi-class classification on massive information through incremental learning.

The invention provides a method for quickly recognizing a speaker and belongs to speaker recognition methods. The method comprises the following steps of: combining a Gaussian mixture model, and taking the supervector of the Gaussian mixture model as the feature parameter of the speaker; taking the supervector of the Gaussian mixture model as input, designing a one-class support vector machine classifier; and training N classifiers corresponding to N speakers, thus obtaining a voice sample of one speaker from one classifier. By utilizing the method, the speaker recognition speed is increased; for every new registered speaker, only one one-class support vector machine classifier is trained for the new speaker, so that the speaker recognition system has good extensibility.

A method for developing a template of a reference document, such as a bank note and using that template to validate other test documents. The method involves using images of a plurality of reference documents, such as banknotes and segmenting each image in a like manner into a plurality of segments. Each segment is then classified using a one-class classifier to determine a reference classification parameter. These parameters are used to define a threshold reference classification parameter. Optionally, the step of segmenting is optimized to determine a preferred number of segments, and a preferred combination of these for use in a subsequent validation process. Preferably, this is done using a stochastic algorithm such as a genetic algorithm.

The invention discloses an on-line diagnostic method for an abnormal energy consumption branch of a building. The method comprises the following steps of: (1) circularly detecting all branches of an entire building according to a preset branch detection sequence; (2) determining detection reference time, and if a branch is detected for the first time, reading energy consumption information of thebranch from an energy consumption information management database and reconstructing to a phase space according to a phase space theory; and (3) providing branch abnormal alarm information corresponding to the energy consumption data when the detected energy consumption data is abnormal. The invention provides the method for performing data mining from a great capacity of energy consumption information of the building branch and discovering abnormal energy consumption data; and by the method, dynamic modeling and real-time abnormal data judgment can be realized and an adaptive diagnostic model is established by scrolling correction. The method solves the problem of processing the non-linear abnormal data detection by other methods based on a phase space reconstruction theory and classification technology by using a kernel function.

Disclosed herein is an anomaly detection method for link-state routing protocols, a link-state routing protocol providing for link-state update (LSU) messages to be exchanged between nodes in a packet-based network, wherein each link-state update message includes link-state advertisement (LSA) message(s) each having a respective header. The method comprises monitoring the link-state advertisement messages exchanged in the network, extracting and forming respective feature vectors with the values in the fields of the headers of the monitored link-state advertisement messages, and detecting an anomaly related to routing based on the feature vectors. In particular, detecting an anomaly related to routing includes feeding the feature vectors to a machine learning system, conveniently a one-class classifier, preferably a one-class support vector machine (OC-SVM).

A system, method and computer program product provides a solution to a class of categorization problems using a semi-supervised clustering approach, the method employing performing a Soft Seeded k-means algorithm, which makes effective use of the side information provided by seeds with a wide range of confidence levels, even when they do not provide complete coverage of the pre-defined categories. The semi-supervised clustering is achieved through the introductions of a seed re-assignment penalty measure and model selection measure.

An ECU identifying apparatus and a method of controlling the same are provided. The ECU identifying apparatus includes a communication interface configured to transmit and receive CAN data to and from a plurality of ECUs and a processor configured to measure a power signal of the received CAN data, wherein the processor generates a multi-class classifier with respect to each of the plurality of ECUs and a one-class classifier with respect to all ECUs, acquires identification information of the received CAN data, acquires a signal of a predetermined area from the measured power signal, calculates a predetermined attribute value based on the signal of the predetermined area which is acquired, identifies an ECU based on the identification information of the CAN data which is acquired and the calculated predetermined attribute value, and determines whether an attack is made based on the identified ECU.

Rewarding a first classification of purchasers of digital media is provided. Such a classification may include initial purchasers (e.g., purchasers of new items) and secondary purchasers (e.g., purchasers of previously used items). A digital medium is played on a digital device and a notification is sent to a server. The server identifies the user of the digital device based on a unique device identifier associated with the digital device and a stored association of that digital device with the user. The user is determined to belong to the first classification of purchasers of the digital medium based on a unique identifier associated with the digital medium. The user, belonging to the first classification of purchasers, is provided with access to a reward that is not generally accessible to users belonging to a second classification. Users that are identified as belonging to the second classification may be offered a chance to upgrade their classification and access the reward.

A method for monitoring machine conditions provides additional information using a one-class classifier in which an evaluation function is learned. In the method, a distance is determined from an anomaly measurement x to at boundary of a region R1 containing all acceptable measurements. The distance is used as a measure of the extent of the anomaly. The distance is found by searching along a line from the anomaly to a closest acceptable measurement within the region R1.

The invention discloses a multi-view human facial image gender identification method and a device, wherein the method comprises the following steps: a step of matching by classifiers and a step of identifying human facial image gender. The step of matching by the classifiers comprises the following steps: S11, obtaining a plurality of sample human facial images from a human facial image library; S12, extracting sample characteristics of all sample human facial images and fuzzy matching the first classifier according to the sample characteristics; and S13, testing the sample characteristics by the first classifier and screening a target characteristic, and precisely matching the second classifier according to the target characteristic. The step of identifying human facial image gender comprises the following steps: S21, collecting video images of the target characteristic containing human faces to be detected; S22, extracting the target characteristic of the human faces to be detected; and S23, processing the target characteristic by the second classifier and identifying the human facial gender. The method and device disclosed by the invention can improve the identification rate of human facial images collected from different views.

The invention provides a turbulence target detection method based on BP neural network multi-class classification. The method comprises the step of according to a turbulence echo mode, under differentsignal to noise ratio conditions, generating echo amplitude sequences, and training sets and test sets of a BP neural network formed by corresponding strength grades so as to allow the trained BP neural network to divide the turbulence intensity into multiple grades, thereby achieving turbulence detection. According to the invention, an experience formulas and parameterization models are not required, by use of the multi-class classification function of the neural network, the echo amplitude sequences of a weather target radar are used as input data of the training set of the neural network;the turbulence intensity grades are used as output data of the training set; the relation between the radar echoes and turbulence intensity can be effectively determined only through learning of massecho data; and an objective that turbulence detection is achieved by use of the neural network to classify the turbulence intensities can be achieved.

A method includes: a) preparing as training data a sample set that contains a plurality of samples belonging to a first class and a plurality of samples belonging to a second class; b) generating, by performing discriminant analysis on the sample set, a first discriminant function having a high classification characteristic for the first class and a second discriminant function having a high classification characteristic for the second class; c) by classifying the sample set using the first and second discriminant functions, isolating any sample whose classification results by the first and second discriminant functions do not match; d) forming a new sample set by grouping together any sample thus isolated, and repeating b) and c) by using the new sample set; and e) causing d) to stop when the number of samples each of whose classification results do not match in c) has decreased to or below a predetermined value.

The time taken to learn a model from training examples is often unacceptable. For instance, training language understanding models with Adaboost or SVMs can take weeks or longer based on numerous training examples. Parallelization thought the use of multiple processors may improve learning speed. The invention describes effective methods to distributed multiclass classification learning on several processors. These methods are applicable to multiclass models where the training process may be split into training of independent binary classifiers.

The invention provides a remote-sensing image one-class classification method based on one-class normalization. One-class samples are selected from a remote-sensing image, one-class normalization transformation is carried out, a suprasphere is determined with an original point as a center for one-class classification after normalization transformation is finished, and the sample away from the original point by the distance smaller than the radius of the suprasphere is the positive class. Compared with other one-class classification technologies, the remote-sensing image one-class classification method based on one-class normalization has the advantages that no parameter is needed, namely one-class classification is carried out directly without setting any parameters, the effect is good, and performance is stable.

The invention relates to a wearable fall detection method and system based on hierarchical classification. The method comprises a step of collecting the daily behavior data of a user, a step of carrying out synthesis and filtering processing on the daily behavior data to generate original data, a step of using a sliding window mechanism to extract the time domain feature of the original data, generating samples, and combining the samples to form a sample set, a step of using a first layer first class model to identify each sample in the sample set, and sending an identified result to a weighted two-class classification model of a second layer, a step of allowing the weighted two-class classification model of the second layer to be responsible for weighted distribution processing, generating a weighted fall sample, and sending the weighted fall sample to a rule two-class classification model of a third layer, and a step of allowing the rule two-class classification model of a third layer to judge whether a user falls according to a condition that whether the weighted fall sample is in accordance with a fall rule or not. Through the above method, the accurate judgment of a user fall behavior is realized.

The invention discloses an SAR target identification method based on a scattering point and a K-center one-class classifier, and mainly overcomes the defects that in the prior art the difference of a target from a clutter false-alarm is not revealed in the view of the nature of radar imaging, and the total identification accuracy rate is low. According to the technical scheme, the SAR target identification method comprises the following steps: 1) detecting the clutter false-alarm of an SAR image, and extracting sections; 2) selecting sections with real targets from the extracted sections, thereby forming a training sample; 3) extracting a scattering point matrix from the training sample according to a scattering point model, and performing amplitude 2-norm normalization; 4) performing K-center clustering on the scattering point matrix of the training sample, thereby obtaining a cluster center; 5) calculating an identification threshold Thr of the K-center one-class classifier; 6) calculating the minimum bidirectional Hausdorff distance of a testing sample and the cluster center; 7) judging whether the testing sample is a target or not according to the distance. Through the adoption of the SAR target identification method, the false-alarm rate of clutter false-alarm is reduced, the total identification accuracy rate is effectively increased, and the SAR target identification method is applicable to artificial targets such as vehicles with remarkable strong scattering point distribution characteristics in SAR images.

A system, method and computer program product provides a solution to a class of categorization problems using a semi-supervised clustering approach, the method employing performing a Soft Seeded k-means algorithm, which makes effective use of the side information provided by seeds with a wide range of confidence levels, even when they do not provide complete coverage of the pre-defined categories. The semi-supervised clustering is achieved through the introductions of a seed re-assignment penalty measure and model selection measure.

Optimizing multi-class image classification by leveraging patch-based features extracted from weakly supervised images to train classifiers is described. A corpus of images associated with a set of labels may be received. One or more patches may be extracted from individual images in the corpus. Patch-based features may be extracted from the one or more patches and patch representations may be extracted from individual patches of the one or more patches. The patches may be arranged into clusters based at least in part on the patch-based features. At least some of the individual patches may be removed from individual clusters based at least in part on determined similarity values that are representative of similarity between the individual patches. The system may train classifiers based in part on patch-based features extracted from patches in the refined clusters. The classifiers may be used to accurately and efficiently classify new images.

The invention discloses a natural resource vector-oriented surface coverage change statistical processing method and a storage medium. According to the method, administrative division is taken as a statistical unit, data of a plurality of different historical periods are matched, precisely spliced and cut, statistical analysis of different ground class classification codes is carried out after surface covering layer space superposition, and finally different types of statistical data reports are selected according to needs and are output. Statistical analysis is carried out by taking administrative division as a statistical unit; the volume of each data processing task is reduced; parallel processing resources of an existing computer can be utilized; A large amount of historical data oriented to natural resource vector surface coverage is processed at the same time, the processing speed and the processing efficiency are improved, query can be carried out in a mode of combination of administrative division codes and ground class distribution codes, and classification of different administrative divisions, ground class classification, classification area summarization and multi-dimensional extraction are achieved.