Abstract: An image processing method of extracting the points of a path following a threadlike structure in an image (IP) formed by a grid of Potential points. A first processing step (4) implements a first path-tracking operation using a front marching technique denoted Filiation Front Marching Technique (FFM) for supplying at least one First Track of the threadlike structure, formed by succeeding points denoted Fathers and Children, by marching a Front of points forwards from a fixed Start point (A) to a fixed End point (B). A second processing step (5) implements a second path-tracking operation using the Filiation Front Marching Technique for supplying a Best Path from one First Track by back propagating the Front starting at the End Point and going through already determined Children and Fathers until the Start Point is reached.

Abstract: A system and method facilitating pattern recognition is provided. The invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). The feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. The pattern recognition system can be trained utilizing a calculated cross entropy error. The calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.

Abstract: A Self-Designing Intelligent Signal Processing System Capable of Evolutional Learning for Classification/Recognition of One and Multidimensional Signals is described which classifies data by an evolutionary learning environment that develops the features and algorithms that are best suited for the recognition problem under consideration. The System adaptively learns what data need to be extracted in order to recognize the given pattern with the least amount of processing. The System decides what features need to be selected for classification and/or recognition to fit a certain structure that leads to the least amount of processing according to the nature of the given data. The System disclosed herein is capable of recognizing an enormously large number of patterns with a high accuracy.

Abstract: The system can be used for the automatic analysis of images, including a matrix of spots, such as images of DNA microarrays after hybridization. The system can be associated—and preferably integrated in a single monolithic component implementing VLSI CMOS technology—to a sensor for acquiring the images. The system includes a circuit for processing the signals corresponding to the images, configured according to a cellular neural network (CNN) architecture for the parallel analogue processing of signals.

Abstract: A method for monitoring fabrication processes of finely structured surfaces in a semiconductor fabrication includes the steps of providing reference signatures of finely structured surfaces, measuring at least one signature of a test specimen surface, comparing the measured signature with the reference signatures, and classifying the test specimen surface by using the comparison results, wherein the measurement of the reference signatures is carried out by measuring the local distribution and/or intensity distribution of diffraction images on production prototypes having a specified quality. The classification is preferably carried out here with a neural network having a learning capability and/or a fuzzy logic. Furthermore, a device for carrying out the method is provided.

Abstract: Aspects of the present invention can be configured to demultiplex an asynchronously multiplexed video signal, which comprises images from a number of different cameras. Image matching techniques are used to assign input images with states. After a period, the number of states will generally equal the number of input cameras. The states may be modeled through any number of techniques, such as histogram analysis, clustering, and hidden Markov model analysis. Input images are assigned to states, and the input images are output as being associated with the states. Zone surveillance may be performed on a series of images from one or more of the states. Any events that occur can be distinguished and reported.

Abstract: A method and system for categorizing non-textual subject data, such as digital images, content-based data and meta-data to determine outcomes of classification tasks. The meta-data is indicative of the operational conditions of a recording device during the capturing of the content-based data. For example, the non-textual subject data may be a digital image captured by a digital camera, and the meta-data may include automatic gain setting, film speed, shutter speed, aperture/lens index, focusing distance, date and time, and flash/no flash operation. The subject image is tagged with selected classifiers by subjecting the image to a series of classification tasks utilizing both content-based data and meta data to determine classifiers associated with the subject image.

Abstract: A person having one or more implanted or surface attached electrodes navigates a window environment. Signals from the electrodes are attached to input circuits in a control circuit which includes a sequential digital control circuit. The control circuit is coupled to a processor running window environment software and having a display. While observing the display, the person is able to activate signals by activating or relaxing a muscle or by mental processes such as thought, to thereby navigate the software environment.

Abstract: A learning type image classification apparatus is capable of classifying a plurality of images flexibly. A region clipping mode selection section is capable of selecting a mode of clipping regions from images from a plurality of candidates. The region clipping execution section clips regions from images in a mode selected by the selection section of the region clipping mode.

Abstract: A method and system for image processing, in conjunction with classification of images between natural pictures and synthetic graphics, using SGLD texture (e.g., variance, bias, skewness, and fitness), color discreteness (e.g., R_L, R_U, and R_V normalized histograms), or edge features (e.g., pixels per detected edge, horizontal edges, and vertical edges) is provided. In another embodiment, a picture/graphics classifier using combinations of SGLD texture, color discreteness, and edge features is provided. In still another embodiment, a “soft” image classifier using combinations of two (2) or more SGLD texture, color discreteness, and edge features is provided. The “soft” classifier uses image features to classify areas of an input image in picture, graphics, or fuzzy classes.

Abstract: A multi-neural net imaging apparatus and method for classification of image elements, such as biological particles. The multi-net structure utilizes subgroups of particle features to partition the decision space by an attribute or physical characteristic of the particle and/or by individual and group particle classification that includes an unknown category. Preprocessing and post processing enables heuristic information to be included as part of the decision making process. Preprocessing classifies particles as artifacts based on certain physical characteristics. Post processing enables the use of contextual information either available from other sources or gleaned from the actual decision making process to further process the probability factors and enhance the decisions.

Abstract: Image data of optically acquired input images (1) are processed for emphasizing at least two object classes. Each pixel is subjected to a rough classification (10) based on first criteria that determine whether or not a pixel is relevant for an object recognition. A reduced image (11) is formed from the relevant pixels and irrelevant pixels are omitted. The reduced image (11) is filtered (20) for forming at least two correlated filtered images (21, 22, 23) based on second criteria. Classified images (31A, 32A, 33A) are formed from the filtered images by classifiers that work in accordance with predetermined rules. Weighting factors are allocated to each object class. The classified images are merged in accordance with an algorithm to make a combined global evaluation for each object class. The global evaluation decides, based on the merged images (41A, 41B, 41C), for each pixel whether the respective pixel belongs to an object class and if so to which object class.

Abstract: Temporal processing for realtime human vision system behavioral modeling is added at the input of a spatial realtime human vision system behavioral modeling algorithm. The temporal processing includes a linear and a non-linear temporal filter in series in each of a reference channel and a test channel, the input to the reference channel being a reference image signal and the input to the test channel being a test image signal that is an impaired version of the reference image signal. The non-linear temporal filter emulates a process with neural attack and decay to compensate for a shift in peak sensitivity and for frequency doubling in a spatio-temporal sensitivity function. The linear temporal filter accounts for the remaining subtleties in the spatio-temporal sensitivity function.

Abstract: The present invention may be embodied in a method, and in a related apparatus, for classifying a feature in an image frame. In the method, an original image frame having an array of pixels is transformed using Gabor-wavelet transformations to generate a transformed image frame. Each pixel of the transformed image is associated with a respective pixel of the original image frame and is represented by a predetermined number of wavelet component values. A pixel of the transformed image frame associated with the feature is selected for analysis. A neural network is provided that has an output and a predetermined number of inputs. Each input of the neural network is associated with a respective wavelet component value of the selected pixel. The neural network classifies the local feature based on the wavelet component values, and indicates a class of the feature at an output of the neural network.

Abstract: A method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. The method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite Radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output.

Abstract: There is disclosed a method for converting color image data inputted from an image input apparatus to a color space which is not dependent on the apparatus and/or lighting with a high precision, the method comprising: setting a plurality of sets of a subject as a main constituting element of an image inputted from the image input apparatus, means for estimating color data which is not dependent on the apparatus and/or the lighting for each set, judging whether or not each pixel of the image data inputted from the image input apparatus belongs to the set of the subject, and selecting the estimating means based on a judgment result to estimate the color data which is not dependent on the apparatus and/or the lighting; learning a distribution of a specified object for each set in the color space after color conversion during color conversion of the image, performing tentative color conversion from an input image signal, and using the signal after the tentative color conversion to judge the set to which each pixe

Abstract: An image processing method includes the steps of: separating a luminance signal from a color signal of an original image, and multiscale retinex processing only the luminance signal; and producing an image based on the multiscale retinex processed luminance signal and the color signal of the original image.

Abstract: An application server component of an apparatus in one example receives one or more notifications of one or more open communication sessions from one or more first communication devices. The application server component sends one or more of the one or more notifications of one or more of the one or more open communication sessions to one or more second communication devices. Upon a request by a communication device of the one or more second communication devices to join an open communication session of the one or more of the one or more open communication sessions, the application server component initiates a connection of the communication device to the open communication session.

Abstract: A technology of the present invention is capable of objectively judging an ability of a proofreader who proofreads a digitized document by use OCR programs. A method of managing an ability of a proofreader who proofreads an electronic document generated from a recognition target document by executing a character auto recognition program, comprises a step of estimating a character count of potential mis-recognized characters contained in the electronic document, a step of detecting a mis-recognized character discover count as a mis-recognized character count with which the proofreader discovers the mis-recognized characters in the electronic document, a step of detecting a processing time spent for proofreading the electronic document, and a step of calculating a score relative to a proofreader ability based on a ratio of the potential mis-recognized character count to the mis-recognized character discover count per unit time.

Abstract: A method for identifying a type of a mammographic view for a digital mammography image. The method comprises the steps of: identifying two or more candidate view types; identifying at least one feature capable of distinguishing between the two or more candidate view types; determining the feature for the digital mammography image; and corresponding the determined feature of the digital mammography image with one of the two or more candidate view types to identify the type of a mammographic view of the digital mammography image in accordance with the correspondence.

Abstract: A method of training an artificial neural network (ANN) involves receiving a likelihood distribution map as a teacher image, receiving a training image, moving a local window across sub-regions of the training image to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to the ANN so that it provides output pixel values that are compared to output pixel values of corresponding teacher image pixel values to determine an error, and training the ANN to reduce the error. A method of detecting a target structure in an image involves scanning a local window across sub-regions of the image by moving the local window for each sub-region so as to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to an ANN so that it provides respective output pixel values that represent likelihoods that respective image pixels are part of a target structure, the output pixel values collectively constituting a likelihood distribution map.

Abstract: Computerized aesthetic judgment of images is disclosed. In one embodiment, a computer-implemented method inputs a training set of images, where each image has a corresponding set of aesthetic scores. The method trains a classifier based on the training set, and outputs the classifier. An image can then be input into the classifier, such that an aesthetic score for the image is generated by the classifier and output. Furthermore, recommendations can be generated to improve the aesthetic score for the image, which are also output.

Abstract: Apparatus and method for recognizing counterfeit currency are disclosed. The apparatus comprises capture unit for capturing a digital image of a currency bill to be recognized, a data storage for storing parameters, features, weights, and a feature identification instruction, feature-capturing unit for capturing features of the bill, neural network recognition unit for comparing the features of the bill with that of an authentic bill by performing an back propagation algorithm and using a plastic perception network as a training kernel, and output means for displaying a comparison result.

Abstract: A method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. The color of a standard is expressed as color values. The neural network includes an input layer having nodes for receiving input data related to paint bases. Weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. An output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. The data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. The paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines.

Abstract: A spatial image processor neural network for processing image data to discriminate between first and second spatial configurations of component objects includes a photo transducer input array for converting an input image to pixel data and sending the data to a localized gain network (LGN) module, a parallel memory processor and neuron array for receiving the pixel data and processing the pixel data into component recognition vectors and chaotic oscillators for receiving the recognition vectors and sending feedback data to the LGN module as attention activations. The network further includes a temporal spatial retina for receiving both the pixel data and temporal feedback activations and generating temporal spatial vectors, which are processed by a temporal parallel processor into temporal component recognition vectors. A spatial recognition vector array receives the temporal component recognition vectors and forms an object representation of the first configuration of component objects.

Abstract: Disclosed herein are systems and methods for facilitating the usage of an online workforce to remotely monitor security-sensitive sites and report potential security breaches.

Abstract: A method and system improve the detection of abnormalities, such as lung nodules, in radiological images using digital image processing and artificial neural network techniques. The detection method and system use a nodule phantom for matching in order to enhance the efficiency in detection. The detection method and system use spherical parameters to characterize true nodules, thus enabling detection of the nodules in the mediastinum. The detection method and system use a multi-layer back-propagation neural network architecture not only for the classification of lung nodules but also for the integration of detection results from different classifiers. In addition, this method and system improve the detection efficiency by recommending the ranking of true nodules and several false positive nodules prior to the training of the neural network classifier. The method and system use image segmentation to remove regions outside the chest in order to reduce the false positives outside the chest region.

Abstract: A method and apparatus for classifying a plurality of elements in images, where electronic images of a field of view containing elements are formed. Each of the elements has a plurality of features. A first subgroup of the plurality of features from the images of the plurality of elements are extracted and processed to segregate the plurality of elements into first and second groups. A classification class only for each of the elements in the first group is determined by selecting and processing a second subgroup of the extracted features to determine a physical characteristic of the element, and by selecting and processing a third subgroup of the extracted features in response to the determined physical characteristic to determine a classification class of the element. The second group of elements bypasses the determination of classification class.

Abstract: An input picture signal having noise added through a transmission path or the like is supplied. Noise is removed from the input picture by a class categorizing adaptive process in which predictive coefficients are pre-learnt and decided for each class. A class corresponding to a noise component contained in the input picture is decided. Predictive coefficients of the class and the values of pixels of frames containing a considered pixel of the input picture signal are linearly combined. Thus, predictive pixel values are generated, which are free of noise. When motion of a considered pixel is detected and pixels that are used to decide a class and pixels that are used for a predictive calculation are compensated corresponding to the motion, noise accurately corresponding to a noise component can be removed from the input picture signal.

Abstract: A method, system, and computer program product of selecting a set of training images for a massive training artificial neural network (MTANN). The method comprises selecting the set of training images from a set of domain images; training the MTANN with the set of training images; applying a plurality of images from the set of domain images to the trained MTANN to obtain a corresponding plurality of scores; and determining the set of training images based on the plurality of images, the corresponding plurality of scores, and the set of training images. The method is useful for the reduction of false positives in computerized detection of abnormalities in medical images. In particular, the MTAAN can be used for the detection of lung nodules in low-dose CT (LDCT). The MTANN consists of a modified multilayer artificial neural network capable of operating on image data directly.

Abstract: A content-based classification system is provided that detects the presence of object images within a frame and determines the path, or trajectory, of each object image through multiple frames of a video segment. In a preferred embodiment, face objects and text objects are used for identifying distinguishing object trajectories. A combination of face, text, and other trajectory information is used in a preferred embodiment of this invention to classify each segment of a video sequence. In one embodiment, a hierarchical information structure is utilized to enhance the classification process. At the upper, video, information layer, the parameters used for the classification process include, for example, the number of object trajectories of each type within the segment, an average duration for each object type trajectory, and so on. At the lowest, model, information layer, the parameters include, for example, the type, color, and size of the object image corresponding to each object trajectory.

Abstract: A method for recognizing an object image comprises the steps of extracting a candidate for a predetermined object image from an overall image, and making a judgment as to whether the extracted candidate for the predetermined object image is or is not the predetermined object image. The candidate for the predetermined object image is extracted by causing the center point of a view window, which has a predetermined size, to travel to the position of the candidate for the predetermined object image, and determining an extraction area in accordance with the size and/or the shape of the candidate for the predetermined object image, the center point of the view window being taken as a reference during the determination of the extraction area.

Abstract: A computer-implemented method and apparatus for information organization, wherein atomic information can be both static and dynamic, but the compound information (e.g., associations, groupings, sets, etc.) of such atoms always remain dynamic. Unless otherwise directed, a compound information entity is always dynamically determined and generated. This determination is based on the processing of a defined condition, wherein all atoms qualifying the condition are included in the compound. This dynamic determination eliminates the need to “update” the compound, when atoms and/or compounds common to two or more compounds are changed. Further, each information compound can be dynamically generated based on an existing definition for that compound.

Abstract: Techniques for converting images defined by halftone bitmaps to continuous tone (CT) representations can better preserve geometric information from the original halftone bitmaps, enhancing the accuracy of CT proofs produced by lower resolution proofers such as inkjet and electrophotographic devices. The conversion techniques may involve the application of different conversion processes to text/linework and image regions of the bitmaps. For example, the conversion process for image regions may involve application of bandwidth limitation to remove halftone dot structures prior to downsampling. On the contrary, the conversion process for text/linework regions may exclude bandwidth limitation in order to preserve sharpness of text and linework.

Abstract: A method and apparatus for color matching are provided, in which paint recipe neural networks are utilized. The color of a standard is expressed as color values. The neural network includes an input layer having nodes for receiving input data related to paint bases. Weighted connections connect to the nodes of the input layer and have coefficients for weighting the input data. An output layer having nodes are either directly or indirectly connected to the weighted connections and generates output data related to color values. The data to the input layer and the data from the output layer are interrelated through the neural network's nonlinear relationship. The paint color matching neural network can be used for, but not limited to, color formula correction, matching from scratch, effect pigment identification, selection of targets for color tools, searching existing formulas for the closest match, identification of formula mistakes, development of color tolerances and enhancing conversion routines.

Abstract: A method for transforming an input variable having a plurality of binary input positions into an m-position output variable in accordance with a predeterminable transformation specification. The transformation is effected in the form of a plurality of input positions or intermediate positions combined group-wise in parallel and sequential partial transformations with intermediate positions being created according to predeterminable, discrete partial assignment specifications. In addition, an arrangement for transforming an n-position input variable having a plurality of binary input positions into an m-position output variable according to a predeterminable transformation specification in the form of a network.

Abstract: In a method for controlling a processing apparatus, an error value between an input value of the processing apparatus for processing a subject to be processed, and a measurement value obtained by measuring the subject being processed is obtained. A correction value is computed for correcting the input value of the processing apparatus in the direction of decreasing the error value, and the values are managed as processing data to be utilized in computing a next correction value. Previous processing data having a history identical to that of the subject loaded to the processing apparatus is searched, and a current bias correction value is predicted from a plurality of most recent correction values having the identical history. Also, a current random correction value is predicted by means of a neural network on the basis of a plurality of most recent random correction values. The predicted bias correction value is summed with the random correction value as a current correction value of the processing apparatus.

Abstract: Methods and systems of the present invention may be used to recognize digital image data arranged in rows and columns. Exemplary embodiments may include a feature extractor for extracting feature information from data representing the rows and columns of the digital image data, a feature compressor for compressing the extracted feature information, and a neural network for classifying the digital image data from the compressed, extracted feature information.

Abstract: A quadrature mirror filter is applied to decompose an image into at least two sub-images each having a different resolution. These decomposed sub-images pass through self-organizing map neural networks for performing a non-supervisory classification learning. In a testing stage, the recognition process is performed from sub-images having a lower resolution. If the image can not be identified in this low resolution, the possible candidates are further recognized in a higher level of resolution.

Abstract: A process for the processing of signals by which signal S is optimally verified for association with objects of a desired class Zk (wherein (with 1<=k<=number of different classes) and differentiated from undesired objects. One or more classes ZA of undesired objects, and beyond this, as an additional class, a rejection class R is defined, to which all signals are assigned which are not clearly assignable to one of classes Zk or ZA. Herein the decision criteria under which a signal S is assigned to the reject class R is the comparison of the output value Preject with the adjustable threshold value t, which is provided by a classification algorithm.

Abstract: An electro-optical method and apparatus for evaluating the dimensions of any protrusion from the threshold of the fabric surface is achieved by bending any length of fabric over a rotating roller so that the contoured area of the protrusion body above the surface can be visualized. The image of the contour as seen by a digital camera is processed statistically and then by a neural network to yield an integrated picture of the fabric protrusions. The grading results of pilling are well correlated to the visual method of pilling evaluation.

Abstract: A data processing apparatus includes a determining unit capable of determining, in accordance with a plurality of data to be extracted from said input data, irregular intervals among said plurality of data to be extracted from said input data, an extracting unit for extracting the plurality of data, which correspond to output data of interest to be predicted, from said input data in accordance with the determination result by said determining unit, and a predicting unit for finding predicted values of the output data of interest based on the plurality of data extracted by said extracting unit.

Abstract: The model is a third generation neurosimulator. It has a plurality of areas whose functions can be identified with the functions of the areas of the dorsal and ventral path of the visual cortex of the human brain. Feedback is provided between different areas during processing. There is additionally provided competition for attention between different features and/or different spatial regions. The model is very flexibly suitable for image processing. It simulates natural human image processing and explains many experimentally observed phenomena.

Abstract: In the prediction of a resin pattern, images of a plurality of sample resins corresponding to pattern material information are captured first. Then, from among the captured images, a plurality of images are selected on the basis of the pattern material data, and are combined to synthesize an image for output. The image synthesis and output process is repeated by varying the pattern material data until the output image shows a desired resin pattern, and thereby the pattern material data for obtaining a resin having the desired resin pattern is determined. In the prediction of a resin color, the color of a resin molding is predicted from color material data, and the predicted image is output. The prediction and output process is repeated by varying the color material data until the output image matches an image of a resin molding having a desired color, and thereby the color material data for obtaining the resin molding having the desired color is determined.

Abstract: Detecting video phenomena, such as fire in an aircraft cargo bay, includes receiving a plurality of video images from a plurality of sources, compensating the images to provide enhanced images, extracting features from the enhanced images, and combining the features from the plurality of sources to detect the video phenomena. The plurality of sources may include cameras having a sensitivity of between 400 nm and 1000 nm and/or may include cameras having a sensitivity of between 7 and 14 micrometers. Extracting features may include determining an energy indicator for each of a subset of the plurality of frames. Detecting video phenomena may also include comparing energy indicators for each of the subset of the plurality of frames to a reference frame.

Abstract: A method of training an artificial neural network (ANN) involves receiving a likelihood distribution map as a teacher image, receiving a training image, moving a local window across sub-regions of the training image to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to the ANN so that it provides output pixel values that are compared to output pixel values of corresponding teacher image pixel values to determine an error, and training the ANN to reduce the error. A method of detecting a target structure in an image involves scanning a local window across sub-regions of the image by moving the local window for each sub-region so as to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to an ANN so that it provides respective output pixel values that represent likelihoods that respective image pixels are part of a target structure, the output pixel values collectively constituting a likelihood distribution map.

Abstract: Embodiments of the present invention comprise methods and systems for automatically adjusting images to conform to preference data.

Abstract: The present invention describes a partial independent component analysis (PICA) technique for blindly separating partially independent and/or gaussian-like sources from mixed observations over an informative index subspace, which allows various applications in independent component imaging. The present invention estimates a demixing matrix using only the independent and/or nongaussian portion of the observations. Specifically, rather than using all the data points which give rise to a large separation error, a subset of the data points is identified such that the partial source profiles defined over such a subset are statistically independent and/or nongaussian. The present invention describes a complete implementation of such a technique, whose steps and parameters may be achieved and estimated using an information theoretic-based neural computational algorithm.

Abstract: A text mining program is provided that allows a user to perform text mining operations, such as: information retrieval, term and document visualization, term and document clustering, term and document classification, summarization of individual documents and groups of documents, and document cross-referencing. This is accomplished by representing the text of a document collection using subspace transformations. This subspace transformation representation is performed by: constructing a term frequency matrix of the term frequencies for each of the documents, transforming the term frequencies for statistical purposes, and projecting the documents or the terms into a lower dimensional subspace. As the document collection is updated, the subspace is dynamically updated to reflect the new document collection.

Abstract: An artificial neural network (ANN) based system that is adapted to process an input pattern to generate an output pattern related thereto having a different number of components than the input pattern. The system (26) is comprised of an ANN (27) and a memory (28), such as a DRAM memory, that are serially connected. The input pattern (23) is applied to a processor (22), where it can be processed or not (the most general case), before it is applied to the ANN and stored therein as a prototype (if learned). A category is associated with each stored prototype. The processor computes the coefficients that allow the determination of the estimated values of the output pattern, these coefficients are the components of a so-called intermediate pattern (24). Assuming the ANN has already learned a number of input patterns, when a new input pattern is presented to the ANN in the recognition phase, the category of the closest prototype is output therefrom and is used as a pointer to the memory.