Abstract: The present invention relates to a learning system. The learning system comprises a sensory and perception module, a cognitive module, and an execution module. The sensory and perception module is configured to receive and process external sensory input from an external world and extract sensory-specific features from the external sensory input. The cognitive module is configured to receive the sensory-specific features and identify a current context based on the sensory-specific features. Based on the current context and features, the cognitive module learns, constructs, or recalls a set of action plans and evaluates the set of action plans against any previously known action plans in a related context. Based on the evaluation, the cognitive module selects the most appropriate action plan given the current context. The execution module is configured to carry out the action plan.

Abstract: A method for characterizing, detecting and predicting an event of interest, a target event, based on temporal patterns useful for predicting a probable occurrence of the target event is disclosed. Measurable events and their features are defined and quantized into event classes. Temporal series of the event classes are analyzed, and preliminary prediction rules established by analyzing temporal patterns of the event classes that precede an occurrence of the target event using a sliding time window. The quality of the preliminary prediction rules is evaluated and parameters thereof are optimized by using a defined fitness function, thereby defining finalized prediction rules. The finalized prediction rules are then made available for application on temporal series of the event classes to forecast a probable occurrence of the target event.

Abstract: A video detection and monitoring method and apparatus utilizes an application-specific object based segmentation and recognition system for locating and tracking an object of interest within a number of sequential frames of data collected by a video camera or similar device. One embodiment includes a background modeling and object segmentation module to isolate from a current frame at least one segment of the current frame containing a possible object of interest, and a classification module adapted to determine whether or not any segment of the output from the background modeling apparatus includes an object of interest and to characterize any such segment as an object segment. An object segment tracking apparatus is adapted to track the location within a current frame of any object segment and to determine a projected location of the object segment in a subsequent frame.

Abstract: A method for enhancing the quality of a digital image by using a single user-defined parameter. A virtual image is created based on the single user-defined parameter and the original digital image. An adaptive contrast enhancement algorithm operates on a logarithmically compressed version of the virtual image to produce adaptive contrast values for each pixel in the virtual image. A dynamic range adjustment algorithm is used to generate logarithmic enhanced pixels based on the adaptive contrast values and the pixels of the logarithmically compressed version of the virtual image. The logarithmic enhanced pixels are exponentially expanded and scaled to produce a compensated digital image.

Abstract: A method for enhancing the quality of a digital image by using a single user-defined parameter. A virtual image is created based on the single user-defined parameter and the original digital image. An adaptive contrast enhancement algorithm operates on a logarithmically compressed version of the virtual image to produce adaptive contrast values for each pixel in the virtual image. A dynamic range adjustment algorithm is used to generate logarithmic enhanced pixels based on the adaptive contrast values and the pixels of the logarithmically compressed version of the virtual image. The logarithmic enhanced pixels are exponentially expanded and scaled to produce a compensated digital image.

Abstract: An apparatus and method is disclosed for providing application specific multi-dimensional information to an application running on a user computing device, wherein at least one dimension of the information is a category, from a plurality of member documents electronically extracted from a library of electronically searchable documents, which may comprise an application specific multidimensional information extractor adapted to extract occurrences of prospective representations of dimensions of application specific multidimensional information from the member documents, and to extract occurrences of non-application specific multidimensional information from the member documents; and, an encoder adapted to encode the occurrences of prospective dimensions of application specific multidimensional information and non-application specific multidimensional information contained in member documents according to a dimension specific coded representation of each dimension of application specific multidimensional inform

Abstract: A system, method, and apparatus for signal characterization, estimation, and prediction comprising an integrated search algorithm that cooperatively optimizes several data mining sub-tasks, the integrated search algorithm including a machine learning model, and the method comprising processing the data for data embedding, data embedding the processed data for searching for patterns, extracting time and frequency patterns, and training the model to represent learned patterns for signal characterization, estimation, and prediction.

Abstract: An apparatus and method provides application specific multidimensional information to an application running on a user computing device from a plurality of member documents electronically extracted from a library of electronically searchable documents. An information extractor is adapted to extract occurrences of prospective representations of dimensions of application specific multidimensional information and occurrences of non-application specific multidimensional information from the member documents. Also, an encoder is adapted to encode the occurrences of prospective dimensions of application specific and non-application specific multidimensional information contained in member documents. A member document identifier determines document formatting and decides whether to proceed with further processing. An information verification unit optionally verifies the extraction of application specific multidimensional information from the member documents.

Abstract: The present invention provides an improved method and system for training an on-line fuzzy inference network to generate a rule base, and a rule base generated thereby. Tuning and applying a learning rule to the fuzzy rules generated by the fuzzy inference network in such as manner as to divorce the performance of the network from the number of input dimensions allows the present invention to adapt a fuzzy inference network such as a SONFIN to be effective for the classification of high-dimensional data in problems requiring the use of a high number of dimensions such as occupant recognition in vehicles, weather forecasting, and economic forecasting.

Abstract: Described is an active learning system for fingerprinting an object identified in an image frame. The active learning system comprises a flow-based object segmentation module for segmenting a potential object candidate from a video sequence, a fixed-basis function decomposition module using Haar wavelets to extract a relevant feature set from the potential object candidate, a static classifier for initial classification of the potential object candidate, an incremental learning module for predicting a general class of the potential object candidate, an oriented localized filter module to extract features from the potential object candidate, and a learning-feature graph-fingerprinting module configured to receive the features and build a fingerprint of the object for tracking the object.

Abstract: A plurality of image chips (202) (over 100), each of the chips containing the same, known target of interest, such as, for example an M109 tank are presented to the system for training. Each image chip of the known target is slightly different than the next, showing the known target at different aspect angles and rotation with respect to the moving platform acquiring the image chip. The system extract multiple features of the known target from the plurality of image chips (202) presented for storage and analysis, or training. These features distinguish a known target of interest from the nearest similar target to the M109 tank, for example a Caterpillar D7 bulldozer. These features are stored for use during unknown target identification. When an unknown target chip is presented, the recognition algorithm relies on the features stored during training to attempt to identify the target.

Abstract: A vision-based system for automatically detecting the type of object within a specified area, such as the type of occupant within a vehicle. Determination of the type of occupant can then be used to determine whether an airbag deployment system should be enabled or not. The system extracts different features from images captured by image sensors. These features are then processed by classification algorithms to produce occupant class confidences for various occupant types. The occupant class confidences are then fused and processed to determine the type of occupant. In a preferred embodiment, image features derived from image edges, motion, and range are used. Classification algorithms may be implemented by using trained C5 decision trees, trained Nonlinear Discriminant Analysis networks, Hausdorff template matching and trained Fuzzy Aggregate Networks.

Abstract: A method for enhancing the quality of a digital image by using a single user-defined parameter. A virtual image is created based on the single user-defined parameter and the original digital image. An adaptive contrast enhancement algorithm operates on a logarithmically compressed version of the virtual image to produce adaptive contrast values for each pixel in the virtual image. A dynamic range adjustment algorithm is used to generate logarithmic enhanced pixels based on the adaptive contrast values and the pixels of the logarithmically compressed version of the virtual image. The logarithmic enhanced pixels are exponentially expanded and scaled to produce a compensated digital image.

Abstract: Object detection and tracking operations on images that may be performed independently are presented. A detection module receives images, extracts edges in horizontal and vertical directions, and generates an edge map where object-regions are ranked by their immediacy. Filters remove attached edges and ensure regions have proper proportions/size. The regions are tested using a geometric constraint to ensure proper shape, and are fit with best-fit rectangles, which are merged or deleted depending on their relationships. Remaining rectangles are objects. A tracking module receives images in which objects are detected and uses Euclidean distance/edge density criterion to match objects. If objects are matched, clustering determines whether the object is new; if not, a sum-of-squared-difference in intensity test locates matching objects.

Abstract: A vision-based system for automatically detecting the type of object within a specified area, such as the type of occupant within a vehicle is presented. The type of occupant can then be used to determine whether an airbag deployment system should be enabled or not. The system extracts different features, including wavelet features and/or a disparity map from images captured by image sensors. These features are then processed by classification algorithms to produce class confidences for various occupant types. The occupant class confidences are fused and processed to determine occupant type. In a preferred embodiment, image features from image edges, wavelet features, and disparity are used. Various classification algorithms may be implemented to classify the object. Use of the disparity map and/or wavelet features provides greater computational efficiency.

Abstract: An computer vision method and system for recognizing and tracking occupants in a fixed space under variable illumination. The system utilizes a camera to capture an initial image of the unoccupied fixed space and subsequently captures images of the occupied fixed space. The edge maps of the current estimate of the unoccupied fixed space including illumination variations and the occupied fixed space are computed. The edge map of the current estimate of the unoccupied fixed space is then subtracted from the edge map of the occupied fixed space to yield a residual edge map, which is then processed to extract the image of the occupant. At least one equivalent rectangle is then computed from the two-dimensional moments of the image of the occupant. The equivalent rectangles are then used to determine the occupant type and position and to track changes in real-time. This method and system is generally designed for use with automobile safety systems such as “smart” airbags.

Abstract: An apparatus and method are disclosed for electronically extracting application specific multidimensional information from a library of electronically searchable documents, wherein at least one dimension of the information is a category, which may comprise an automatic document miner in communication with the contents of the library and adapted to electronically extract relevant documents from the library; an E-Space filter creator adapted to create from the extracted relevant documents a category specific representation of the extracted relevant documents comprising the E-Space filter; a document selector adapted to utilize the E-Space filter to separate the extracted relevant documents into member documents and non-member documents and to discard the non-member documents; and an application specific multidimensional information extractor adapted to extract occurrences of application specific multidimensional information from the member documents.

Abstract: Apparatus and method for providing application specific multi-dimensional information to an application running on a user computing device, wherein at least one dimension of the information is a category, from a plurality of member documents electronically extracted from a library of electronically searchable documents, may comprise an application specific multidimensional information extractor to extract occurrences of prospective representations of dimensions of application specific multidimensional information and to extract occurrences of non-application specific multidimensional information; and, an encoder adapted to encode the occurrences of prospective dimensions of application specific multidimensional information and non-application specific multidimensional information contained in member documents according to a dimension specific coded representation of each dimension of application specific multidimensional information and a non-application specific coded representation of each non-application s

Abstract: The present invention provides an improved method and system for training an on-line fuzzy inference network to generate a rule base, and a rule base generated thereby. Tuning and applying a learning rule to the fuzzy rules generated by the fuzzy inference network in such as manner as to divorce the performance of the network from the number of input dimensions allows the present invention to adapt a fuzzy inference network such as a SONFIN to be effective for the classification of high-dimensional data in problems requiring the use of a high number of dimensions such as occupant recognition in vehicles, weather forecasting, and economic forecasting.

Abstract: A boosting and pruning system and method for utilizing a plurality of neural networks, preferably those based on adaptive resonance theory (ART), in order to increase pattern classification accuracy is presented. The method utilizes a plurality of N randomly ordered copies of the input data, which is passed to a plurality of sets of booster networks. Each of the plurality of N randomly ordered copies of the input data is divided into a plurality of portions, preferably with an equal allocation of the data corresponding to each class for which recognition is desired. The plurality of portions is used to train the set of booster networks. The rules generated by the set of booster networks are then pruned in an intra-booster pruning step, which uses a pair-wise Fuzzy AND operation to determine rule overlap and to eliminate rules which are sufficiently similar. This process results in a set of intra-booster pruned booster networks.