Abstract: Systems and methods are provided through which a graphic image is classified in terms of being natural versus computer generated, or being a scientific slide presentation versus a comic image. The image is classified by extracting appropriate feature(s) from the image, and using the feature(s) to determine, within a predetermined degree of accuracy, the graphic classification of the image. The classification determination uses a trained model. The trained model is created by using machine learning algorithms such as Neural Networks, Support Vector Machines, and Learning Vector Quantizations. Subsequently, the trained model is used to classify a group of images of unknown classification automatically.

Abstract: A procedure for fast training and evaluation image classification systems using support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw image data by means of a set of m linear functions defined on the k-dimensional raw input data, and are used for image classification, including facial recognition tasks.

Abstract: An apparatus and method for time synchronization of a plurality of multimedia streams are described. In one embodiment, the method includes the concurrent capture of multimedia data via a plurality of multimedia platforms. During the concurrent capture of the multimedia data, each of the multimedia platforms receives a synchronization signal from a synchronization generation unit. In response, each platform processes a received synchronization signal to generate a common reference clock signal among each of the platforms. Once the common clock signal is generated, each of the platforms synchronizes captured multimedia data to form multimedia stream data according to the common reference clock signal. As such, the plurality of multimedia platforms are able to perform collaborative signal processing tasks of multimedia streams, including, for example, array signal processing algorithms.

Abstract: A method, machine readable medium, and system are disclosed. In one embodiment the method comprises generating an acoustic signal from an actuator of a first computing device, receiving the acoustic signal with a sensor of a second computing device, receiving the acoustic signal with a sensor of a third computing device, generating an estimate of a difference between the amount of time required for the acoustic signal to travel from the actuator of the first computing device to the sensor of the second computing device and the amount of time required for the acoustic signal to travel from the actuator of the first computing device to the sensor of the third computing device, wherein the sensors and actuator are unsynchronized, and computing, based on the estimated difference in time, a physical location of at least one of the said sensors and actuator.

Abstract: A cache management operation. In one embodiment, a first recall value for a first unit of data is generated, a second recall value for a second unit of data is generated, and the first and second recall values are compared. The unit of data having the higher recall value is stored in a first section of a storage device. The unit of data having the lower recall value is stored in a second section of a storage device. A greater amount of compression is performed on the unit of data having the lower recall value.

Abstract: A television signal substitution system that replaces known video segments such as advertisements with selected replacement advertisements. Fingerprint data of known advertisements can be stored in a fingerprint database. When new fingerprint data is available, the fingerprint data can be automatically or manually transmitted to subscribers. Various techniques can be used to identify advertisements based on the fingerprint data.

Abstract: A tree classifier may include a number of stages. Some stages may include monolithic classifiers, and other stages may be split into two or more classifiers.

Abstract: An apparatus and method for time synchronization of a plurality of multimedia streams are described. In one embodiment, the method includes the concurrent capture of multimedia data via a plurality of multimedia platforms. During the concurrent capture of the multimedia data, each of the multimedia platforms receives a synchronization signal from a synchronization generation unit. In response, each platform processes a received synchronization signal to generate a common reference clock signal among each of the platforms. Once the common clock signal is generated, each of the platforms synchronizes captured multimedia data to form multimedia stream data according to the common reference clock signal. As such, the plurality of multimedia platforms are able to perform collaborative signal processing tasks of multimedia streams, including, for example, array signal processing algorithms.

Abstract: A procedure for fast training and evaluation of support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw input data by means of a set of m linear functions defined on the k-dimensional raw input data. Training uses a one-time precomputation on the linear transform matrix in order to allow training on an equivalent training set with vector size k instead of m, given a great computational benefit in case of m>>k. A similar pre-computation is used during evaluation of SVMs, so that the raw input data vector can be used instead of the derived linear feature vector.

Abstract: A procedure for fast training and evaluation image classification systems using support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw image data by means of a set of m linear functions defined on the k-dimensional raw input data, and are used for image classification, including facial recognition tasks.

Abstract: A cache management operation. In one embodiment, a first recall value for a first unit of data is generated, a second recall value for a second unit of data is generated, and the first and second recall values are compared. The unit of data having the higher recall value is stored in a first section of a storage device. The unit of data having the lower recall value is stored in a second section of a storage device. A greater amount of compression is performed on the unit of data having the lower recall value.

Abstract: In some embodiments, the invention includes a method for locating text in digital images. The method includes scaling a digital image into images of multiple resolutions and classifying whether pixels in the multiple resolutions are part of a text region. The method also includes integrating scales to create a scale integration saliency map and using the saliency map to create initial text bounding boxes through expanding the boxes from rectangles of pixels including at least one pixel to include groups of at least one pixel adjacent to the rectangles, wherein the groups have a particular relationship to a first threshold. The initial text bounding boxes are consolidated.

Abstract: A method to obtain a resource among nodes. In one embodiment, a first node receives an inquiry for a resource from a second node. The first node generates a cost evaluation of obtaining the resource and the first node transmits the cost evaluation to the second node for use in selecting where to obtain the resource.

Abstract: A method and apparatus to support a first peer node receiving an inquiry for data from a second peer node. In one embodiment, the first peer node transcodes the data before transmitting the data to the second peer node, wherein the transcoding includes converting the data into a format that can be processed by the second peer node, and transmitting the data to the second peer node in a transport specification as requested by the second peer node.

Abstract: Systems and methods are provided through which a graphic image is classified in terms of being natural versus computer generated, or being a scientific slide presentation versus a comic image. The image is classified by extracting appropriate feature(s) from the image, and using the feature(s) to determine, within a predetermined degree of accuracy, the graphic classification of the image.