289 results about "Statistical learning" patented technology

A prosody generation apparatus capable of suppressing distortion that occurs when generating prosodic patterns and therefore generating a natural prosody is provided. A prosody changing point extraction unit in this apparatus extracts a prosody changing point located at the beginning and the ending of a sentence, the beginning and the ending of a breath group, an accent position and the like. A selection rule and a transformation rule of a prosodic pattern including the prosody changing point is generated by means of a statistical or learning technique and the thus generate rules are stored in a representative prosodic pattern selection rule table and a transformation rule table beforehand. A pattern selection unit selects a representative prosodic pattern from the representative prosodic pattern selection rule table according to the selection rule. A prosody generation unit transforms the selected pattern according to the transformation rule and carries out interpolation with respect to portions other than the prosody changing points so as to generate prosody as a whole.

The invention relates to a movement human abnormal behavior identification method based on template matching, which mainly comprises the steps of: video image acquisition and behavior characteristic extraction. The movement human abnormal behavior identification method is a mode identification technology based on statistical learning of samples. The movement of a human is analyzed and comprehended by using a computer vision technology, the behavior identification is directly carried out based on geometric calculation of a movement region and recording and alarming are carried out; the Gaussian filtering denoising and the neighborhood denoising are combined for realizing the denoising, thereby improving the independent analysis property and the intelligent monitoring capacity of an intelligent monitoring system, achieving higher identification accuracy for abnormal behaviors, effectively removing the complex background and the noise of a vision acquired image, and improving the efficiency and the robustness of the detection algorithm. The invention has simple modeling, simple algorithm and accurate detection, can be widely applied to occasions of banks, museums and the like, and is also helpful to improve the safety monitoring level of public occasions.

Mechanical systems wear or change over time. Data collected over a system's life can be input to statistical learning models to predict this wear/change. Previous work by the inventors trained a flexible empirical regression model at a fixed point of wear, and then applied it independently at time points over the life of an engine to predict wear. The embodiment disclosed herein relates those wear predictions over time using a time evolution model. The time evolution model is sequentially updated with new data, and effectively tunes the empirical model for each engine. The combined model predicts wear with dramatically reduced variability. The benefit of reduced variability is that engine wear is more evident, and it is possible to detect operational anomalies more quickly. In addition to tracking wear, the model is also used as the basis for a Bayesian approach to monitor for sudden changes and reject outliers, and adapt the model after these events.

Systems, methods and apparatus for automatic signal detection with temporal feature extraction in an RF environment are disclosed. An apparatus learns the RF environment in a predetermined period based on statistical learning techniques, thereby creating learning data. A knowledge map is formed based on the learning data. The apparatus automatically extracts temporal features of the RF environment from the knowledge map. A real-time spectral sweep is scrubbed against the knowledge map. The apparatus is operable to detect a signal in the RF environment, which has a low power level or is a narrowband signal buried in a wideband signal, and which cannot be identified otherwise.

The invention discloses a method for controlling game characters based on a human face analyzing technology. The method comprises the following steps: preprocessing an image obtained from an image input device simply and effectively; detecting the human face and positioning key points by a statistical learning method; and analyzing the positioning result to obtain human face gesture and expression information, and mapping the information to corresponding game control commands to achieve the real-time control function of the faces of the game characters. Using human face information of a player to control the game characters extends the interaction modes of the prior games by only shooting a picture of the human face of the player by a camera, analyzing the state and the expression information of the picture in a space, and converting the analyzed result into the control commands of the game characters. A video detection method of the invention is real-time, robust and easy to realize and operate. The method can enable game players to perform interactive operations in a more natural and more intellectual new mode, such as adopting self head gestures, face expressions and the like to endow games with more interactions and immersion.

Systems, methods and apparatus are disclosed for automatic signal detection in an RF environment. An apparatus comprises at least one receiver and at least one processor coupled with at least one memory. The apparatus is at the edge of a communication network. The apparatus sweeps and learns the RF environment in a predetermined period based on statistical learning techniques, thereby creating learning data. The apparatus forms a knowledge map based on the learning data, scrubs a real-time spectral sweep against the knowledge map, and creates impressions on the RF environment based on a machine learning algorithm. The apparatus is operable to detect at least one signal in the RF environment.

An innovative optical system and method is disclosed for analyzing and uniquely identifying high-order refractive indices samples in a diverse population of nearly identical samples. The system and method are particularly suitable for ultra-fine materials having similar color, shape and features which are difficult to identify through conventional chemical, physical, electrical or optical methods due to a lack of distinguishing features. The invention discloses a uniquely configured optical system which employs polarized sample light passing through a full wave compensation plate, a linear polarizer analyzer and a quarter wave retardation plate for producing vivid color bi-refringence pattern images which uniquely identify high-order refractive indices samples in a diverse population of nearly visually identical samples. The resultant patterns display very subtle differences between species which are frequently indiscernable by conventional microscopy methods. When these images are analyzed with a trainable with a statistical learning model, such as a soft-margin support vector machine with a Gaussian RBF kernel, good discrimination is obtained on a feature set extracted from Gabor wavelet transforms and color distribution angles of each image. By constraining the Gabor center frequencies to be low, the resulting system can attain classification accuracy in excess of 90% for vertically oriented images, and in excess of 80% for randomly oriented images.

The invention discloses a method and a device for judging the type of a webpage. The method comprises the following steps: carrying out rule matching in a prestored rule table on the basis of the URL of a webpage to be judged, wherein the rule list comprises a plurality of rule records for judging the type of the webpage; if the rule matching is successful, obtaining the type of the webpage to be judged according to the successfully matched rules; if the rule matching is unsuccessful, extracting predetermined features from the URL and/or HTML source codes of the webpage to be judged, and utilizing a classifier to classify the type of the webpage to be judged so as to obtain the type of the webpage to be judged on the basis of a feature vector composed of features selected from the extracted predetermined features. In the scheme of the invention, the invention can combines the advantages of a rule recognition scheme and a recognition scheme based on statistical learning, and can realize to judge the types of various webpage such as blogs, forums, news and the like.

Provided is a method for modeling the cost of XML as well as relational operators. As with traditional relational cost estimation, a set of system catalog statistics that summarizes the XML data is exploited; however, the novel use of a set of simple path statistics is also proposed. A new statistical learning technique called transform regression is utilized instead of detailed analytical models to predict the overall cost of an operator. Additionally, a query optimizer in a database is enabled to be self-tuning, automatically adapting to changes over time in the query workload and in the system environment.

A computer-implemented method for non-rigid multi-modal registration, includes providing trained data corresponding to previously registered images captured by different modalities, receiving two or more images captured by the different modalities, and registering the two or more images according to context information determined based on the trained data.

A novel statistical learning procedure that can be applied to many machine-learning applications is presented. Although this boosting learning procedure is described with respect to its applicability to face detection, it can be applied to speech recognition, text classification, image retrieval, document routing, online learning and medical diagnosis classification problems.

The present invention provides a rapid and effective image retrieval method under a large-scale data background, and belongs to the technical field of computer vision, statistical learning and pattern recognition. First, local features of an image are extracted by using deep convolutional neural networks after transfer learning is performed in a specific task data set, and further the extracted image local features are modeled by using spatial mean pooling and covariance descriptors. The present invention provides an improved maximum likelihood estimation method to robustly estimate the high-dimensional covariance descriptors. The final image expression is obtained by weighted fusion of a spatial mean pooling model and the covariance descriptors. A low-rank measure learning method based on maximum margin subspace is provided to compare image expressions of two images. Dimensions of the image expressions are reduced to improve image matching efficiency, and image matching accuracy can be improved according to prior information of the specific task data set.

The invention discloses a multi-feature fusion identification algorithm used for human face comparison, which relates to the human face recognition computer technology research and is used for solving the problems of low accuracy and poor actual application effect caused by an existing portrait identification system which does not form a learning sample library and does not change and adjust an actual application weight of an algorithm. The multi-feature fusion identification algorithm comprises a template database and the learning sample library; the template database configures one or more algorithms for different business modules to perform modeling, comparison and application capability identification; algorithm weights, scores and result combination rules are set, and a final identification score is obtained and stored in the learning sample library; and the learning sample library performs statistical learning for a training sample library and hit of an actual application comparison result, and dynamically adjusts the algorithm weights. By recording an application result of each calculation, the actual application effect is subjected to statistics, and application weights of various algorithms are adjusted according to the effect, so that unified scheduling and cooperative running of the various algorithms can be realized and the advantages of the various algorithms can be exerted.

A learning apparatus includes an image generator, a feature point extractor, a feature value calculator, and a classifier generator. The image generator generates, from an input image, images having differing scale coefficients. The feature point extractor extracts feature points from each image generated by the image generator. The feature value calculator calculates feature values for the feature points by filtering the feature points using a predetermined filter. The classifier generator generates one or more classifiers for detecting a predetermined target object from an image by means of statistical learning using the feature values.

The invention discloses a method for detecting human body parts by performing parallel statistical learning based on three-dimensional depth image information. Specific to the problems of complex deformation, difficulty to describe, and the like, of human body parts (head, hands and feet), a novel feature, namely universal feature, for reflecting diversity of the human body parts, is constructed. A parallel statistical learning method is utilized to select effective and sufficient novel features and form a parallel cascaded classifier, so as to perform real-time efficient detection on the human body parts.

The invention discloses a statistical learning model based gate position allocation method. The statistical learning model based gate position allocation method comprises the steps of adopting a prior probability prediction model, producing a take-off time difference value probability set, a take-off time difference value probability set and a landing time difference value probability set according to the historical flying situation of a certain flight, predicting arrival time probability distribution and parking apron airside leisure degree of the flight and accordingly performing flight gate position allocation. The statistical learning model based gate position allocation method is based on airside allocation, facilitates adjustment on only close gate positions of flights and shortens walking distances of passengers. In addition, overall evaluation on the using situation of gate positions is facilitated, allocation can be performed according to the planned landing time probability of the flights based on probability allocation, the gate position allocation accuracy is improved, gate position adjustment times caused by flight delay is decreased, and meanwhile the satisfaction degree of the passengers is improved. The utilization rate of the gate positions is comprehensively improved, and reasonable allocation of gate position resources is ensured.

The invention relates to the field of information processing technology and provides a method and a device for pushing information. The method includes: acquiring historical behavior information of a user from social data sources; dividing the historical behavior information into one or more documents related to the behavior information of the user according to preset rules; using a statistical learning method to acquire models according to the documents; generating push information according to the models, and sending the push information to a client of the corresponding user. Due to the fact that the push information is generated on the basis of the historical behavior information of the user, accuracy of the push information can be increased effectively.

The invention discloses a lake and reservoir algal bloom predicating method based on multielement nonstationary time series analysis and neural network and support vector machine compensation, and belongs to the technical field of water quality monitoring. The method comprises the steps of characteristic factor nonstationary time series modeling, error influence factor kernel principal component analysis, neural network error modeling according to the situation of large sample data, support vector machine error modeling according to the situation of small sample data, final error compensation and predicating result obtaining. The problems that existing algal bloom predication precision is not high, and predication is hard to carry out according to the small sample data are solved, the description of the algal bloom forming process corresponds to reality better, and the result of algal bloom modeling predication is more accurate. The advantage compensation of a time series analysis method suitable for linear system modeling and a statistical learning method suitable for nonlinear system modeling is achieved, and the algal bloom predication accuracy is improved.

The invention puts forward a vehicle video speed measurement method based on a multi-viewpoint camera, and is used for solving the problem of big error of vehicle speed detection since the measurementdistance of a single camera is unstable. The method comprises the following steps that: calibrating a plurality of cameras; carrying out rectification on an image coordinate and a world coordinate; according to multiple viewpoints, positioning a vehicle target; and according to positioning data, measuring the movement distance and the movement speed of the vehicle. By use of the method, through multi-camera calibration, monitoring picture collection and a method based on statistical learning, the target is accurately positioned, vehicle speed is calculated to realize the speed measurement ofthe video, the positioning accuracy of the target vehicle in video collection data is improved, in addition, an inspection missing problem in the single-camera target detection is avoided and reduced,the calculation amount of the image coordinate and the world coordinate is small, the real-time detection of video monitoring can be realized, and an alarm is given. The method is high in measurementand wide in applicable ranges, a situation that single camera detection is not integral can be avoided, and a severe environment is coped.

The invention discloses a construction method of a real-time bus passenger analysis system. The construction method includes the steps: performing pretreatment analysis for historical bus data according to data processing rules and historical data analysis of bi-directional bus card swiping passengers based on a Spark cloud computing platform; acquiring getting off behaviors of various stations of the passengers by the aid of historical data of bi-directional getting on behaviors of the passengers, and constructing getting on and getting off station distribution of bus passengers of routes by a statistic learning method; acquiring a data analysis computing method of passenger density of intervals by the aid of the getting on behaviors of the passengers; pre-processing bus data in real time by the aid of Spark, and dynamically generating bus station lists by the aid of bus stop records; feeding back background data processing results to foregrounds in real time by the aid of Spark Streaming technology, and reading and writing dynamic bus data at high speed by the aid of Redis; and analyzing the bus data in real time, and computing passenger density according to analysis results of the historical data.