Abstract: Embodiments of the present invention relate to object tracking in video. In an embodiment, a computer-implemented method tracks an object in a frame of a video. An adaptive term value is determined based on an adaptive model and at least a portion of the frame. A pose constraint value is determined based on a pose model and at least a portion the frame. An alignment confidence score is determined based on an alignment model and at least a portion the frame. Based on the adaptive term value, the pose constraint value, and the alignment confidence score, an energy value is determined. Based on the energy value, a resultant tracking state is determined. The resultant tracking state defines a likely position of the object in the frame given the object's likely position in a set of previous frames in the video.

Abstract: A system identifies an image and determines whether the image contains inappropriate content based on first data associated with the image, second data associated with a document that contains the image or refers to the image, and/or third data associated with a group of documents with which the image is associated.

Abstract: A computer-implemented method of providing personalized route information involves gathering a plurality of past location indicators over time for a wireless client device, determining a future driving objective using the plurality of previously-gathered location indicators, obtaining real-time traffic data for an area proximate to the determined driving objective, and generating a suggested route for the driving objective using the near real-time traffic data.

Abstract: A computer-implemented method of providing personalized route information involves gathering a plurality of past location indicators over time for a wireless client device, determining a future driving objective using the plurality of previously-gathered location indicators, obtaining real-time traffic data for an area proximate to the determined driving objective, and generating a suggested route for the driving objective using the near real-time traffic data.

Abstract: A system identifies an image and determines whether the image contains inappropriate content based on first data associated with the image, second data associated with a document that contains the image or refers to the image, and/or third data associated with a group of documents with which the image is associated.

Abstract: A system analyzes multiple images to identify similar images using histograms, image intensities, edge detectors, or wavelets. The system retrieves labels assigned to the identified similar images and selectively concatenates the extracted labels. The system assigns the concatenated labels to each of the identified similar images and uses the concatenated labels when performing a keyword search of the plurality of images.

Abstract: Messages of a user are ranked based on metrics derived from user actions with regard to the messages. Groupings of messages, such as conversations or a set of messages that are determined to be similar, may also be ranked. When new messages are received, their interest to the user is estimated based on rankings of messages and message groupings to which the new message is related. The new messages may be made more or less prominent based on the estimated interest to the user.

Abstract: One embodiment of the present invention provides a method and a system for building a parallel hybrid spill tree to facilitate parallel nearest-neighbor matching operations. During operation, the system receives a set of objects to be stored in the parallel hybrid spill tree. The system selects a subset of objects from the set of objects, and then uses this subset to create a “top tree.” Each node in the top tree defines an associated partition for the parallel hybrid, spill tree. The system uses the top tree to associate each object in the set of objects with a corresponding partition of the parallel hybrid spill tree. Then, the system builds for each partition of the parallel hybrid spill tree an associated “leaf sub-tree” containing the objects in the partition, with each leaf sub-tree structured as a spill tree.

Abstract: One embodiment of the present invention provides a method and a system for performing a nearest-neighbor matching operation using a parallel hybrid spill tree. During operation, the system receives an object to be compared to a set of objects stored in the parallel hybrid spill tree. The system first searches a “top tree” of the parallel hybrid spill tree to identify a partition that is likely to contain a nearest neighbor of the object. Each node in the top tree defines an associated partition for the parallel hybrid spill tree. The system then searches a “leaf sub-tree” of the parallel hybrid spill tree that corresponds to the associated partition in an attempt to identify the nearest neighbor of the object.

Abstract: A system analyzes multiple images to identify similar images using histograms, image intensities, edge detectors, or wavelets. The system retrieves labels assigned to the identified similar images and selectively concatenates the extracted labels. The system assigns the concatenated labels to each of the identified similar images and uses the concatenated labels when performing a keyword search of the plurality of images.

Abstract: Ink strokes of cursive writing are segmented to make the cursive writing more like print writing, particularly with respect to the number of strokes of a character. A stroke-segmentation module first finds the local extrema points on a stroke of input ink. Then the local extrema points are stepped through, two (or three) at a time. The stroke-segmentation module may compare the three (or four) ink segments that are adjacent to the two (or three) local extrema points to a set of predefined stroke-segmentation patterns to find a closest matching pattern. Strokes are then segmented based on a stroke-segmentation rule that corresponds to the closest matching pattern. Additional stroke segmentation may be performed based on the change of curvature of the segmented ink strokes. Then, a character-recognition module performs character recognition processing by comparing the segmented ink strokes to prototype samples at least some of which have been similarly segmented.

Abstract: Messages of a user are ranked based on metrics derived from user actions with regard to the messages. Groupings of messages, such as conversations or a set of messages that are determined to be similar, may also be ranked. When new messages are received, their interest to the user is estimated based on rankings of messages and message groupings to which the new message is related. The new messages may be made more or less prominent based on the estimated interest to the user.

Abstract: A computer-implemented method of providing personalized route information involves gathering a plurality of past location indicators over time for a wireless client device, determining a future driving objective using the plurality of previously-gathered location indicators, obtaining real-time traffic data for an area proximate to the determined driving objective, and generating a suggested route for the driving objective using the near real-time traffic data.

Abstract: Sets of local, (e.g., online) ads may be generated by obtaining sets of information about (e.g., local) establishments, each set including a business address information and/or a telephone number, (b) determining, for each of the sets, a location using at least one of at least a portion of the business address information and at least a portion of the telephone number, and (c) generating, for each of the sets, an ad that includes targeting information that targets the serving of the ad to queries related to the determined location. A query, including information about a location of a client device, may be accepted and at least one of the generated ads that includes targeting information that targets the location of the client device may be determined.

Abstract: A computer-implemented method for generating results for a client-requested query involves receiving a query produced by a client communication device, generating a result for the query in response to reception of the query, determining one or more predictive follow-up requests before receiving an actual follow-up request from the client device, and initiating retrieval of information associated with the one or more predictive follow-up requests, and transmitting at least part of the result to the client device, and then transmitting to the client device at least part of the information associated with the one or more predictive follow-up requests.

Abstract: A technique for reducing a prototype database for use with handwriting recognizers. The prototype database is divided into multiple spaces and the effect on the error rate for removing each prototype is identified. All prototypes that contribute to the error rate are removed. If further prototype database reduction is desired, the technique identifies and removes one or more prototypes that have the least adverse effect on the error rate across all spaces. The technique continues removing prototypes in this manner until a desired prototype database size has been achieved.

Abstract: Input handwritten characters are classified as print or cursive based upon numerical feature values calculated from the shape of an input character. The feature values are applied to inputs of an artificial neural network which outputs a probability of the input character being print or cursive. If a character is classified as print, it is analyzed by a print character recognizer. If a character is classified as cursive, it is analyzed using a cursive character recognizer. The cursive character recognizer compares the input character to multiple prototype characters using a Dynamic Time Warping (DTW) algorithm.

Abstract: Ink strokes of cursive writing are segmented to make the cursive writing more like print writing, particularly with respect to the number of strokes of a character. A stroke-segmentation module first finds the local extrema points on a stroke of input ink. Then the local extrema points are stepped through, two (or three) at a time. The stroke-segmentation module may compare the three (or four) ink segments that are adjacent to the two (or three) local extrema points to a set of predefined stroke-segmentation patterns to find a closest matching pattern. Strokes are then segmented based on a stroke-segmentation rule that corresponds to the closest matching pattern. Additional stroke segmentation may be performed based on the change of curvature of the segmented ink strokes. Then, a character-recognition module performs character recognition processing by comparing the segmented ink strokes to prototype samples at least some of which have been similarly segmented.

Abstract: A technique for reducing a prototype database for use with handwriting recognizers. The prototype database is divided into multiple spaces and the effect on the error rate for removing each prototype is identified. All prototypes that contribute to the error rate are removed. If further prototype database reduction is desired, the technique identifies and removes one or more prototypes that have the least adverse effect on the error rate across all spaces. The technique continues removing prototypes in this manner until a desired prototype database size has been achieved.

Abstract: A method for detecting a face in an image includes the steps of applying the image to a first classification tool that determines a rotational angle for rotating the image that makes the image most resemble an upright face; rotating the image by the rotational angle determined by the first classification tool; and applying the rotated image to a second classification tool, which determines whether the rotated image represents a frontal face or not.