1286 results about "Feature generation" 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.

A method of managing a fluid or gas reservoir is disclosed which assimilates diverse data having different acquisition time scales and spatial scales of coverage for iteratively producing a reservoir development plan that is used for optimizing an overall performance of a reservoir. The method includes: (a) generating an initial reservoir characterization, (b) from the initial reservoir characterization, generating an initial reservoir development plan, (c) when the reservoir development plan is generated, incrementally advancing and generating a capital spending program, (d) when the capital spending program is generated, monitoring a performance of the reservoir by acquiring high rate monitor data from a first set of data measurements taken in the reservoir and using the high rate monitor data to perform well-regional and field-reservoir evaluations, (e) further monitoring the performance of the reservoir by acquiring low rate monitor data from a second set of data measurements taken in the reservoir, (f) assimilating together the high rate monitor data and the low rate monitor data, (g) from the high rate monitor data and the low rate monitor data, determining when it is necessary to update the initial reservoir development plan to produce a newly updated reservoir development plan, (h) when necessary, updating the initial reservoir development plan to produce the newly updated reservoir development plan, and (i) when the newly updated reservoir development plan is produced, repeating steps (c) through (h). A detailed disclosure is provided herein relating to the step (a) for generating the initial reservoir characterization and the step (b) for generating the initial reservoir development plan.

A capability for prominence-based feature generation and rendering for digital maps is provided. More specifically, embodiments relate to rendering map features such as buildings or landmarks in different rendering styles based on signals for how important a particular feature is to a search context. A search context may be, for example and without limitation, a general view of the map or a user-initiated search request for a particular point of interest or driving directions between different points of interest on the map. For example, the different rendering styles may include, but are not limited to, two-dimensional (2D) footprints, two-and-a-half-dimensional (2.5D) extruded polygons, as will be described further below, and full three-dimensional (3D) models. Furthermore, the style could include color and/or visual texture.

A technique for generating a web site in accordance with received data entry by: presenting a prompt for data entry; receiving a data entry in response to the prompt; presenting at least one additional prompt for data entry that is determined at least in part based upon the received data entry; receiving an additional data entry in response to each of the at least one additional prompts; determining at least one characteristic of at least one web site dimension of the web site based on the data entry; generating a multi-dimensional description of the web site based on the determined characteristics; retrieving web site data according to the generated multi-dimensional description of the web site; and generating the web site based upon the generated multi-dimensional description of the web site and the retrieved web site data.

Techniques, hardware, and software are provided for quantification of extensional features of structures of an imaged subject from image data representing a two-dimensional or three-dimensional image. In one embodiment, stenosis in a blood vessel may be quantified from volumetric image data of the blood vessel. A profile from a selected family of profiles is fit to selected image data. An estimate of cross sectional area of the blood vessel is generated based on the fit profile. Area values may be generated along a longitudinal axis of the vessel, and a one-dimensional profile fit to the generated area values. An objective quantification of stenosis in the vessel may be obtained from the area profile. In some cases, volumetric image data representing the imaged structure may be reformatted to facilitate the quantification, when the structural feature varies along a curvilinear axis. A mask is generated for the structural feature to be quantified based on the volumetric image data. A curve representing the curvilinear axis is determined from the mask by center-finding computations, such as moment calculations, and curve fitting. Image data are generated for oblique cuts at corresponding selected orientations with respect to the curvilinear axis, based on the curve and the volumetric image data. The oblique cuts may be used for suitable further processing, such as image display or quantification.

There is described a system and a method for generating a metatag indicative of an attribute of a person represented in an image, the method comprising: performing a computer-based analysis of the image; identifying a feature related to the attribute of the person in the image using results from the computer-based analysis; generating the metatag based on the identified feature; and associating the metatag to the image, the metatag being thereby related to the attribute of the person.

A system and method that classify driver driving skill based on multiple types of maneuvers. An original features generation processor receives signals identifying driving characteristics of the driver of the vehicle and outputs original features from the signals. A features extraction processor extracts certain features from the original features and outputs transformed features. A features selection processor selects certain ones of the transformed features and outputs final features and a classifier classifies the final features to determine the skill level of the vehicle driver.

A system for gathering data for purposes of analyzing a network. The data is gathered based on values that are passed from node-to-node within processing devices, such as servers, in a network. The values are generated from characteristics obtained from measuring device performance and resource utilization. Values are passed peer-to-peer and successively combined with values at each receiving system so that, ultimately, a value is obtained that reflects the operation of a group of devices. A correlation matrix is maintained to indicate discrepancies in value meanings from different devices. The correlation matrix is used to more accurately combine values to achieve meaningful composite values.

The present invention provides a method for generating a visualizing map of music in accordance with the identifiable features of the music. First, the music would be divided into plural segments, and the length of each segment is preferably identical. After that, an audio analysis is executed to determine the mood types of these segments. Each mood type may be determined by certain parameters, such as tempo value and articulation type. Besides, every mood type corresponds to a certain visualizing expression, and the correspondence can be defined in advance and looked up in a table for example. Eventually, the visualizing map of the music is generated according to the mood types and the distribution of visualizing expressions.

A system and method for comparing a text image and a character string are provided. The method includes embedding a character string into a vectorial space by extracting a set of features from the character string and generating a character string representation based on the extracted features, such as a spatial pyramid bag of characters (SPBOC) representation. A text image is embedded into a vectorial space by extracting a set of features from the text image and generating a text image representation based on the text image extracted features. A compatibility between the text image representation and the character string representation is computed, which includes computing a function of the text image representation and character string representation.

Embodiments of a system and method are described for generating a finance attribute. In one embodiment, the systems and methods retrieve raw tradeline data from a plurality of credit bureaus, retrieve industry code data related to each of the plurality of credit bureaus, determine one or more tradeline leveling characteristics that meet at least one pre-determined threshold, and generate a finance attribute using the selected leveling characteristics.

To accurately classify a query as navigational, thousands of available features are explored, extracted from major commercial search engine results, user Web search click data, query log, and the whole Web's relational content. To obtain the most useful features for navigational query identification, a three level system is used which integrates feature generation, feature integration, and feature selection in a pipeline. Because feature selection plays a key role in classification methodologies, the best feature selection method is coupled with the best classification approach to achieve the best performance for identifying navigational queries. According to one embodiment, linear Support Vector Machine (SVM) is used to rank features and the top ranked features are fed into a Stochastic Gradient Boosting Tree (SGBT) classification method for identifying whether or not a particular query is a navigational query.

Techniques for determining and using a thermodynamic model that characterizes a thermodynamic response of an enclosure conditioned by an HVAC system are disclosed. To determine a thermodynamic model, temperature information when the HVAC system operates in a first state may first be received. A response interval may then be determined where the response interval indicates an estimated time between when the HVAC system begins operating in the first state and when the temperature within the enclosure begins to change in a direction associated with the first state. Weighting factors corresponding to basis functions may then be determined, where the weighted basis functions characterize the temperature trajectory of the enclosure in response to the HVAC system operating in the first state. The basis functions may include a first basis function that is evaluated from a time that the HVAC system begins operating in the first state until a time when the response interval ends, and a second basis function that is evaluated beginning at the time when the response interval ends.

Systems and techniques for scene classification and prediction is provided herein. A first series of image frames of an environment from a moving vehicle may be captured. Traffic participants within the environment may be identified and masked based on a first convolutional neural network (CNN). Temporal classification may be performed to generate a series of image frames associated with temporal predictions based on a scene classification model based on CNNs and a long short-term memory (LSTM) network. Additionally, scene classification may occur based on global average pooling. Feature vectors may be generated based on different series of image frames and a fusion feature vector may be obtained by performing data fusion based on a first feature vector, a second feature vector, a third feature vector, etc. In this way, a behavior predictor may generate a predicted driver behavior based on the fusion feature.

The embodiment of the invention discloses a model training method for synthesizing speaking expressions. Expression characteristics, acoustic characteristics and text characteristics are obtained according to videos containing face action expressions of speakers and corresponding voices. Because the acoustic feature and the text feature are obtained according to the same video, the time interval and the duration of the pronunciation element identified by the text feature are determined according to the acoustic feature. A first corresponding relation is determined according to the time interval and duration of the pronunciation element identified by the text feature and the expression feature, and an expression model is trained according to the first corresponding relation. The expressionmodel can determine different sub-expression characteristics for the same pronunciation element with different durations in the text characteristics; the change patterns of the speaking expressions are added, the speaking expressions generated according to the target expression characteristics is determined by the expression model. The speaking expressions have different change patterns for the same pronunciation element, and therefore the situation that the speaking expressions are excessively unnatural in change is improved to a certain degree.

In one embodiment, a diagnostics system is provided. The diagnostics system include a detection system configured to capture acoustic information and contextual information related to a machine component defect. A computing system is coupled to a processor configured to receive the acoustic information and the contextual information from the detection system, select one or more algorithms based at least in part on the contextual information, and retrieve and execute the one or more algorithms to extract one or more characteristic features of the acoustic information. The processor is further configured to generate an acoustic fingerprint based at least in part on the one or more characteristic features of the acoustic information, such that the one or more characteristic features correspond to the machine component defect.

The present invention describes a road terrain detection system that comprises a method for classifying selected locations in the environment of a vehicle based on sensory input signals such as pixel values of a camera image. The method comprises a high level spatial feature generation for selected locations in the environment called base points. The spatial feature generation of the base points is based on a value-continuous confidence representation that captures visual and physical properties of the environment, generated by so called base classifiers operating on raw sensory data. Consequently, the road terrain detection incorporates both local properties of sensor data and their spatial relationship in a two-step feature extraction process.

Systems and methods which implement a plurality of different imaging signatures in generating an image frame are shown. A first imaging signature may be configured for providing relatively high quality images with respect to subsurface regions of living tissue, for example, whereas a second imaging signature may be configured for providing relatively high quality images with respect to interventional instruments inserted into living tissue at a steep angle. Image sub-frames generated using each such different imaging signature are blended to form a frame of the final image providing a relatively high quality image of various objects within the volume being imaged.

The invention belongs to the technical field of cross-media correlation learning, and particularly discloses a deep cross-mode correlation learning-based image retrieval method for a free-hand sketch.The method comprises three main algorithms of deep multi-mode feature generation, multi-mode correlation learning modeling and similarity sorting optimization. By utilizing a deep learning technology, a depth semantic feature and a depth visual feature are constructed for describing a text tagging part and an image/sketch part in a multi-mode document. Based on a multi-mode document representation, a cross-mode correlation model is built for modeling a whole multi-mode document set, thereby describing correlation among different modes of the multi-mode document. Based on correlation featuresobtained after correlation modeling, retrieval results are sorted and optimized, and color images and texts with the maximum similarity with the queried sketch are returned.