We introduced a novel approach to global sign language recognition by leveraging the capabilities of the Editable Mediator. Traditional methods have often been limited to recognizing sign languages from specific linguistic regions, necessitating ad hoc implementation for multilingual regions. Our method aims to bridge this gap by providing a unified framework for recognizing sign languages in various linguistic areas and promoting global communication. At the core of our system is the Editable Mediator, a mechanism that determines the actual sign meaning from various local hand-shape recognitions. Instead of focusing on specific sign language notations, such as HamNoSys, our approach emphasizes the recognition of common primitive actions shared across different sign languages. These primitive actions are recognized by multiple modules, and their combinations are interpreted by the Editable Mediator to determine the intended sign-language message. This architecture not only simplifies the recognition process, but also offers flexibility. By merely editing the Editable Mediator, our system can adapt to various sign languages worldwide without the need for extensive retraining or ad hoc implementation. This innovation reduces barriers to introducing new sign language systems and promotes a more inclusive global communication platform.

Through technology, it is essential to seamlessly bridge the divide between diverse speaking communities (including the signer (the sign language speaker) community). In order to realize communication that successfully conveys emotions, it is necessary to recognize not only verbal information but also non-verbal information. In the case of signers, there are two main types of behavior: verbal behavior and emotional behavior. This paper presents a sign language recognition method by similarity measure with emotional expression specific to signers. We focus on recognizing the sign language conveying verbal information itself and on recognizing emotional expression. Our method recognizes sign language by time-series similarity measure on a small amount of model data, and at the same time, recognizes emotion expression specific to signers. Our method extracts time-series features of the body, arms, and hands from sign language videos and recognizes them by measuring the similarity of the time-series features. In addition, it recognizes the emotional expressions specific to signers from the time-series features of their faces.

Online learning of a robot's inverse dynamics model for trajectory tracking necessitates an interaction between the robot and its environment to collect training data. This is challenging for physical robots in the real world, especially for humanoids and manipulators due to their large and high dimensional state and action spaces, as a large amount of data must be collected over time. This can put the robot in danger when learning tabula rasa and can also be a time-intensive process especially in a multi-robot setting, where each robot is learning its model from scratch. We propose accelerating learning of the inverse dynamics model for trajectory tracking tasks in this multi-robot setting using knowledge transfer, where robots share and re-use data collected by preexisting robots, in order to speed up learning for new robots. We propose a scheme for collecting a sample of correspondences from the robots for training transfer models, and demonstrate, in simulations, the benefit of knowledge transfer in accelerating online learning of the inverse dynamics model between several robots, including between a low-cost Interbotix PhantomX Pincher arm, and a more expensive and relatively heavier Kuka youBot arm. We show that knowledge transfer can save up to 63% of training time of the youBot arm compared to learning from scratch, and about 58% for the lighter Pincher arm.

This paper firstly analyzes the shortcoming of a self-organizing incremental neural network (SOINN), then proposes a novel online similarity metric and online adaptive kernel density estimator to handle 2 basic problems of unsupervised learning: clustering and density estimation. Our approach is an extension of the standard Gaussian process, online density estimator and SOINN; not only does it fully consider the local structure (the local distribution around each node), but also the learning of parameters is adjusted to make it feasible to online learning. The experiment results show that our method is more accurate and robust to multiscale, multimodal and multivariate data.

With the ongoing development and expansion of communication networks and sensors, massive amounts of data are continuously generated in real time from real environments. Beforehand, prediction of a distribution underlying such data is difficult; furthermore, the data include substantial amounts of noise. These factors make it difficult to estimate probability densities. To handle these issues and massive amounts of data, we propose a nonparametric density estimator that rapidly learns data online and has high robustness. Our approach is an extension of both kernel density estimation (KDE) and a self-organizing incremental neural network (SOINN); therefore, we call our approach KDESOINN. An SOINN provides a clustering method that learns about the given data as networks of prototype of data; more specifically, an SOINN can learn the distribution underlying the given data. Using this information, KDESOINN estimates the probability density function. The results of our experiments show that KDESOINN outperforms or achieves performance comparable to the current state-of-the-art approaches in terms of robustness, learning time, and accuracy.

This work proposes a framework that enables arbitrary robots with unknown kinematics models to imitate human demonstrations to acquire a skill, and reproduce it in real-time. The diversity of robots active in non-laboratory environments is growing constantly, and to this end we present an approach for users to be able to easily teach a skill to a robot with any body configuration. Our proposed method requires a motion trajectory obtained from human demonstrations via a Kinect sensor, which is then projected onto a corresponding human skeleton model. The kinematics mapping between the robot and the human model is learned by employing Local Procrustes Analysis, which enables the transfer of the demonstrated trajectory from the human model to the robot. Finally, the transferred trajectory is modeled using Dynamic Movement Primitives, allowing it to be reproduced in real time. Experiments in simulation on a 4 degree of freedom robot show that our method is able to correctly imitate various skills demonstrated by a human.

This study presents a robotic application of neural associative memory-based control system that imparts online learning and predictive control strategies to a cost-effective quadrotor helicopter, the Parrot AR.Drone 2.0. The control system is extended with to tackle a fundamental and challenging problem for the quadcoptor: hovering control. The proposed system is based on self-organizing incremental neural network that includes an associative memory algorithm. The algorithm can cope with a hierarchical data space and complex time-transition dynamics; it enables online incremental learning from manual control, thereby gradually improve the stability against interference such as drift caused by either mechanical impairment or external excitation. In particular, after continuously learning the associative state-command pair of hovering manoeuvre, the system can execute the command associated with current state. The proposed system is evaluated on a realistic AR.Drone quadcoptor to test its capacity to tackle the hovering control problem. The results demonstrate that for the first time, the proposed system effectively offers a novel approach to quadcoptor application of an associative memory-based neural network by successfully tackling a hover task through iterative on-line learning.

If a robot is expected to perform in the real-world, the robot should recognize objects in such environment using its multimodal sensors in real-time. Traditional multimodal object classification methods focus on recognizing known objects; however, it is impossible to learn all objects that we use. On the other hand, the classification of unknown objects has become a popular topic in image processing. However popular methods have batch algorithms, and there is no method to integrate multimodal classification results with an online algorithm. This study proposes a novel method that estimates multimodal attributes of an unknown object. The method uses an ultra-fast and online learning method based on a STAR-SOINN, which stands for STAtistical Recognition on Self-Organizing and Incremental Neural Network. The results from a comparative experiment show that the recognition accuracy for known objects is higher than a method that naïvely integrates the modalities and a previous method. And this method works very quickly: approximately 1 second to learn one object, and 25 millisecond for a single estimation. We also conducted an experiment to estimate attributes of unknown objects, it could estimate approximately 90% of the attributes for these objects.

Associative learning plays a major role in the formation of the internal dynamic engine of an adaptive system or a cognitive robot. Interaction with the environment can provide a sparse and discrete set of sample correlations of input–output incidences. These incidences of associative data points can provide useful hints for capturing underlying mechanisms that govern the system’s behavioral dynamics. In many approaches to solving this problem, of learning system’s input–output relation, a set of previously prepared data points need to be presented to the learning mechanism, as a training data, before a useful estimations can be obtained. Besides data-coding is usually based on symbolic or nonimplicit representation schemes. In this paper, we propose an incremental learning mechanism that can bootstrap from a state of complete ignorance of any representative sample associations. Besides, the proposed system provides a novel mechanism for data representation in nonlinear manner through the fusion of self-organizing maps and Gaussian receptive fields. Our architecture is based solely on cortically-inspired techniques of coding and learning as: Hebbian plasticity and adaptive populations of neural circuitry for stimuli representation. We define a neural network that captures the problem’s data space components using emergent arrangement of receptive field neurons that self-organize incrementally in response to sparse experiences of system–environment interactions. These learned components are correlated using a process of Hebbian plasticity that relates major components of input space to those of the output space. The viability of the proposed mechanism is demonstrated through multiple experimental setups from real-world regression and robotic arm sensory-motor learning problems.

Clustering is widely used in machine learning, feature extraction, pattern recognition, image analysis, information retrieval, and bioinformatics. Online unsupervised incremental learning is an important branch of data clustering. However, accurately separating high-density overlapped areas in a network has a direct impact on the performance of the clustering algorithm. In this paper, we propose a load-balancing self-organizing incremental neural network (LB-SOINN) to achieve good clustering results and demonstrate that it is more stable than an enhanced SOINN (E-SOINN). LB-SOINN has all the advantages of E-SOINN, such as robustness to noise and online unsupervised incremental learning. It overcomes the shortcomings of the topology structure generated by E-SOINN, such as dependence on the sequence of the input data, and avoids the turbulence that occurs when separating a composite class into subclasses. Furthermore, we also introduce a distance combination framework to obtain good performance for high-dimensional space-clustering tasks. Experiments involving both artificial and real world data sets indicate that LB-SOINN has superior performance in comparison with E-SOINN and other methods.

This paper presents a novel framework for online incremental place recognition in an indoor environment. The framework addresses the scenario in which scene images are gradually obtained during long-term operation in the real-world indoor environment. Multiple users may interact with the classification system and confirm either current or past prediction results; the system then immediately updates itself to improve the classification system. This framework is based on the proposed \emph{n}-value self-organizing and incremental neural network (\emph{n}-SOINN), which has been derived by modifying the original SOINN to be appropriate for use in scene recognition. The evaluation was performed on the standard MIT 67-category indoor scene dataset and shows that the proposed framework achieves the same accuracy as that of the state-of-the-art offline method, while the computation time of the proposed framework is significantly faster and fully incremental update is allowed. Additionally, a small extra set of training samples is incrementally given to the system to simulate the incremental learning situation. The result shows that the proposed framework can leverage such additional samples and achieve the state-of-the-art result.

The paper presents a new online incremental zero-shot learning method for applications in robotics and mobile communications where attribute labeling is obtained via online interaction with users, and where the potential for inconsistency exists. Unique to most previous offline batch learning methods, the proposed method is based on the indirect-attribute-prediction (IAP) model instead of the direct-attribute-prediction (DAP). Using self-organizing and incremental neural networks (SOINN) as the learning mechanism, our method can learn new attributes and update existing attributes in an online incremental manner while retaining as high accuracy as that of the state-of-the-art offline method. Compared to the offline methods, the computation time has also been reduced by more than 99%. Two experiments evaluated two aspects of the proposed method. First, our method clearly outperforms the previous IAP-based offline method in terms of both time and accuracy, and yield approximately the same accuracy as the DAP-based offline method. Second, the proposed method can deal with situations where object attributes are gradually labeled via interaction with many users and where some of them may be incorrect. This scenario is very important for applications in mobile communications and robotics where some objects and attributes may be initially unknown and must be learnt online.

A new approach of unseen place classification in a commercial district is presented. It can classify input scenes into the correct place classes without the needs for sample images of places for training. The number of place classes and their definition are supervised by humans using text information only. A description of individual place classes is obtained from humans as a set of words that are regarded as the disjoint-text-attributes of the unseen place. During classification, our approach determines the number of text-attributes found in an image. Our approach runs in an online incremental manner in the sense that the description of place classes can be updated and a new place class can be added at any time. Our approach can be used, does not require any training dataset, and is available in multiple languages. The evaluation is done by a set of Google Street View images of a shopping area in Japan where both the Japanese and English languages are available. The result shows that the proposed method outperforms the state-of-the-art methods of scene text recognition and standard pattern recognition. The computation is sufficiently fast for real-time application.

Classifying new unseen object classes has become a popular topic of research in the computer-vision and robotics community. Coping with this problem requires determining the attributes shared among objects and transferring them for use in classifying unseen object classes. Nevertheless, most current state-of-the-art methods require a fully offline training process and take a very long time for the batch training process, which renders them inapplicable for use in online applications such as robotics. This study proposes a novel online and incremental approach for learning and transferring the learned attributes in order to classify another disjoint set of image classes. Among three methods proposed in this paper, a method combining those favorable features of a self-organizing incremental neural network (SOINN) and a support vector machine (SVM) achieves the best performance. This method, called the Alt-SOINN-SVM, can run online incrementally, similar to an SOINN, and perform accurate classification, similar to an SVM. An evaluation was performed with 50 classes of an animal with an attributes dataset (>;30,000 images). The results shows that despite the great reduction in both learning time (92.25% reduction) and classification time (99.87% reduction), and possessing the ability for incremental learning on gradually obtained samples, the proposed method offers reasonably good accuracy for classification. Furthermore, the proposed methods are applicable to use with the increasing number of attribute which improves the accuracy gradually and incrementally.

Vision-based mobile robot's simultaneous localization and mapping (SLAM) and navigation has been the source of countless research contributions because of rich sensory output and cost effectiveness of vision sensors. However, existing methods of vision-based SLAM and navigation are not effective for robots to be used in crowded environments such as train stations and shopping malls, because when we extract feature points from an image in crowded environments, many feature points are extracted from not only static objects but also dynamic objects such as humans. By recognizing all such feature points as landmarks, the algorithm collapses and errors occur in map building and self-localization. In this paper, we propose a SLAM and navigation method that is effective even in crowded environments by extracting robust 3D feature points from sequential vision images and odometry. By using the proposed method, we can eliminate unstable feature points extracted from dynamic objects and perform SLAM and navigation stably. We present experiments showing the utility of our approach in crowded environments, including map building and navigation.

This paper presents a novel 3D object recognition method. The proposed objectives are to overcome shortcoming of the appearance-based method, which lacks a spatial relationship between the parts of an object, and those of other 3D model methods, which require complicated computation. The proposed method is based on a voting process. Appearance estimation is introduced in this work in order to deal with the faulty detection problem. We tested our method for object detection and pose estimation, and the results showed that our method improved the average precision and detection time compared to other methods.

This paper presents a fast, online, and incremental solution for an appearance-based loop closure detection problem in an indoor environment. This problem is important in terms of the navigation of mobile robots. Appearance-based Simultaneous Localization And Mapping (SLAM) for a highly dynamic environment, called Position Invariant Robust Feature Navigation (PIRF-Nav), was first proposed by Kawewong et al. in 2010. Their results showed major improvements from other state-of-the-art methods. However, the computational expense of PIRF-Nav is beyond real time, and it consumes a tremendous amount of memory. These two factors hinder the use of PIRF-Nav for mobile robot applications. This study proposed (i) modified PIRF extraction that makes the system more suitable for an indoor environment and (ii) new dictionary management that can eliminate redundant searching and conserve memory consumption. According to the results, our proposed method can finish tasks up to 12 times faster than PIRF-Nav, with only slight percentage decline in a recall, while the precision remains 1. In addition, for a more challenging task, we collected additional data from a crowded university canteen during lunch time. Even in this cluttered environment, our proposed method performs better with real-time processing compared with other methods.

An associative memory (AM) system is proposed to realize incremental learning and temporal sequence learning. The proposed system is constructed with three layer networks: The input layer inputs key vectors, response vectors, and the associative relation between vectors. The memory layer stores input vectors incrementally to corresponding classes. The associative layer builds associative relations between classes. The proposed method can incrementally learn key vectors and response vectors; store and recall both static information and temporal sequence information; and recall information from incomplete or noise-polluted inputs. Experiments using binary data, real-value data, and temporal sequences show that the proposed method works well.

Feature extraction is an effective tool in data mining and machine learning. Many feature extraction methods have been investigated recently. However, few methods can achieve orthogonal components. Non-orthogonal components distort the metric structure of original data space and contain reductant information. In this paper, we propose a feature extraction method, named as incremental orthogonal basis analysis (IOBA), to cope with the challenging endeavors. First, IOBA learns orthogonal components for original data, not only theoretically but also numerically. Second, an innovative way of training data selection is proposed. This selection scheme helps IOBA pick up numerically orthogonal components from training patterns. Third, by designing a self-adaptive threshold technique, no prior knowledge about the number of components is necessary to use IOBA. Moreover, without solving eigenvalue and eigenvector problems, IOBA not only saves large computing loads, but also avoids ill-conditioned problems. Results of experiments show the efficiency of the proposed IOBA.

A new approach for vision-based navigation in mobile robots is presented. Instead of incremental spectral clustering (ISC), which is considered state-of-the-art, the method of self-organizing incremental neural networks (SOINN) is used for visual space clustering. Using SOINN, the number of nodes in the topological map matches well with the environment. The time used for incremental map building is markedly less than that used for ISC. Furthermore, the rate of single image classification is higher than that of ISC.

Associative memory operating in a real environment must perform well in online incremental learning and be robust to noisy data because noisy associative patterns are presented sequentially in a real environment. We propose a novel associative memory that satisfies these requirements. Using the proposed method, new associative pairs that are presented sequentially can be learned accurately without forgetting previously learned patterns. The memory size of the proposed method increases adaptively with learning patterns. Therefore, it suffers neither redundancy nor insufficiency of memory size, even in an environment in which the maximum number of associative pairs to be presented is unknown before learning. Noisy inputs in real environments are classifiable into two types: noise-added original patterns and faultily presented random patterns. The proposed method deals with two types of noise. To our knowledge, no conventional associative memory addresses noise of both types. The proposed associative memory performs as a bidirectional one-to-many or many-to-one associative memory and deals not only with bipolar data, but also with real-valued data. Results demonstrate that the proposed method's features are important for application to an intelligent robot operating in a real environment. The originality of our work consists of two points: employing a growing self-organizing network for an associative memory, and discussing what features are necessary for an associative memory for an intelligent robot and proposing an associative memory that satisfies those requirements.

This paper presents an approach (SOINN-DTW)for recognition of motion (gesture) that is based on the Self-Organizing Incremental Neural Network (SOINN) and Dynamic Time Warping (DTW). Using SOlNN's function of eliminating noise in the input data and representing the distribution of input data, SOINN-PTW method approximates the output distribution of each state in a self-organizing manner corresponding to the input data. The proposed SOINN-DTW method enhances Stochastic Dynamic Time Warping Method (Nakagawa. 1986). Results of experiments show that SOINN-DTW outperforms HMM, CRF. and HCRF in motion data.

This paper presents an on-line, continuously learning mechanism for sequence data. The proposed approach is based on SOINN-DTW method (Okada and Hasegawa, 2007), which is designed for learning of sequence data. It is based on self-organizing incremental neural network (SOINN) and dynamic time warping (DTW). Using SOINNpsilas function represents the topological structure of online input data, the output distribution of each states is represented and adapted in a self-organizing manner corresponding to online input data. Consequently, this method can train a network and estimate parameters of the output distribution using new (on-line) data continuously, based on scarce batch-training data. Through online learning, the recognition accuracy is improved continuously. To confirm the effectiveness of the on-line learning mechanism of SOINN-DTW, we present an extensive set of experiments that demonstrate how our method outperforms the online learning method of HMM in classifying phoneme data.

This paper presents a proposal of a general framework that explicitly models local information and global information in a conditional random field. The proposed method extracts global image features as well as local ones and uses them to predict the scene of the input image. Scene-based top-down information is generated based on the predicted scene. It represents a global spatial configuration of labels and category compatibility over an image. Incorporation of the global information helps to resolve local ambiguities and achieves locally and globally consistent image recognition. In spite of the model's simplicity, the proposed method demonstrates good performance in image labeling of two datasets.

We introduce a prototype-based nearest-neighbor method that is based on a self-organizing incremental neural network (SOINN). It automatically learns the number of prototypes necessary to determine the decision boundary, and it is robust to noisy training data. The experiments with artificial datasets and real-world datasets illustrate the efficiency of the proposed method.

An online semi-supervised active learning algorithm is proposed, which is based on self-organizing incremental neural network (SOINN). The proposed method do not need any priori knowledge such as number of nodes or number of classes; it can automatically learn number of nodes and teacher vectors needed by the current task; It can realize online incremental learning even life-long learning. The experiments for artificial data and real world data show that the proposed method works efficiently.

This paper presents an online semi-supervised clustering algorithm based on a self-organizing incremental neural network (SOINN). Using labeled data and a large amount of unlabeled data, the proposed semi-supervised SOINN (ssSOINN) can automatically learn the topology of input data distribution without any prior knowledge such as the number of nodes or a good network structure; it can subsequently divide the structure into sub-structures as the need arises. Experimental results we obtained for artificial data and real-world data show that the ssSOINN has superior performance for separating data distributions with high-density overlap and that ssSOINN Classifier (S3C) is an efficient classifier.

Associative memory operating in a real environment must perform well on online incremental learning and be robust to noisy data because noisy associative patterns are presented sequentially in a real environment. We propose a novel associative memory that satisfies these needs. Using the proposed method, new associative pairs that are presented sequentially can be learned accurately without forgetting previously learned patterns. The memory size of the proposed method increases adaptively when learning patterns. Therefore, it suffers neither redundancy nor insufficiency of memory size, even in an environment where the maximum number of associative pairs to be presented is unknown before learning. The proposed method deals with two types of noise. To our knowledge, no conventional associative memory deals with both types. The proposed associative memory performs as a bidirectional one-to-many or many-to-one associative memory and deals not only with bipolar data, but also real-valued data. We infer that the proposed method's features are important for application to an intelligent robot operating in a real environment.

This paper presents an unsupervised approach of integrating speech and visual information without using any prepared data. The approach enables a humanoid robot, Incremental Knowledge Robot 1 (IKR1), to learn word meanings. The approach is different from most existing approaches in that the robot learns online from audio-visual input, rather than from stationary data provided in advance. In addition, the robot is capable of learning incrementally, which is considered to be indispensable to lifelong learning. A noise-robust self-organized growing neural network is developed to represent the topological structure of unsupervised online data. We are also developing an active-learning mechanism, called "desire for knowledge," to let the robot select the object for which it possesses the least information for subsequent learning. Experimental results show that the approach raises the efficiency of the learning process. Based on audio and visual data, they construct a mental model for the robot, which forms a basis for constructing IKRI's inner world and builds a bridge connecting the learned concepts with current and past scenes.

We present a new approach for online incremental word acquisition and grammar learning by humanoid robots. Using no data set provided in advance, the proposed system grounds language in a physical context, as mediated by its perceptual capacities. It is carried out using show-and-tell procedures, interacting with its human partner. Moreover, this procedure is open-ended for new words and multiword utterances. These facilities are supported by a self-organizing incremental neural network, which can execute online unsupervised classification and topology learning. Embodied with a mental imagery, the system also learns by both top-down and bottom-up processes, which are the syntactic structures that are contained in utterances. Thereby, it performs simple grammar learning. Under such a multimodal scheme, the robot is able to describe online a given physical context (both static and dynamic) through natural language expressions. It can also perform actions through verbal interactions with its human partner.

This paper describes a face detection technique that enables detection of extremely small faces such as 6 × 6 pixel. This is the first approach to detect very low-resolution faces in the field of face detection. First, we use a conventional AdaBoost-based face detector to show that the face detection rate falls to 39% from 88% as face resolution decreases from 24 × 24 pixels to 6 × 6 pixels. Second, we propose a new face detection method comprising four techniques. It improves the face detection rate from 39% to 73% for 6 × 6 pixel faces of the MIT+CMU frontal face test set. Finally, we applied our method to real world data. By merging results of some frames and judging using a threshold, we show that our method is effective for processing real world data.

This paper proposes a novel framework that integrates bottom-up information and top-down information for scene understanding. Bottom-up information is derived from local features of texture and color. Top-down information is generated from a holistic image context. The information is integrated effectively by extension of the Ising model, which is a simple model of ferromagnetism. Locally and globally consistent image recognition is achieved through an iterative process. The proposed method showed 91.8% accuracy in road-image labeling, which is superior to results obtained using only bottom-up information (81.9%) and the best accuracy obtained using the other method (90.7%).

This study is intended to realize a motion recognition and generation mechanism based on observation. This mechanism, which is based on imitative learning, enables unsupervised incremental learning, recognition, and generation of time-series patterns that are observed directly from motion images. The mechanism segments these patterns into primitives in a self-organized manner using mixture-of-experts (MoE) with a non-monotonous neural network (NMNN). These patterns are expressed as permutations of primitives that are output by the MoE. Applying enhanced dynamic time warping (DTW) method recognizes these permutations of primitives. In addition, we introduce a semi-supervised learning method by applying this mechanism. We confirmed the effectiveness of this mechanism through two experiments using gestures

An on-line learning mechanism is proposed for unsupervised data. Using a similarity threshold and local error based insertion criterion, the system is able to grow incrementally and to accommodate input patterns of online non-stationary data distribution. The definition of a utility parameter -"error-radius" - enables this system to learn the number of nodes needed to solve a task. The usage of a new technique for removing nodes in low probability density regions can separate the clusters with low-density overlaps and dynamically eliminate noise in the input data. Experiment results show that this system can report a reasonable number of clusters and represent the topological structure of unsupervised on-line data with no prior conditions such as a suitable number of nodes or a good initial codebook.

We introduced a new approach to classify 3D texture images based on output selection from two classification methods. For a set of real-world textured images, we found that, using classification based only on texture information, some images obtained under unknown conditions appeared to have similar surfaces and could not be recognized correctly. Another classification method, based on color information, was therefore performed in parallel to solve such a problem. Some knowledge is required to efficiently select outputs from texture-based classification (TBC) and color-based classification (CBC). Reliability tables (RT), created in the pre-testing stage (PTS) of system training, are proposed to obtain that knowledge. Performance is evaluated by employing over 5600 texture images corresponding to 61 real-world surface samples extracted from the Columbia-Utrecht reflectance and texture (CUReT) database K.J. Dana et al., 1999. Our experiments produce better classification results than those in M. Varma and A. Zisserman (2005), which are the best results available to date.

Fractal image coding is a computationally expensive method. This paper does some improvement for traditional no search method and quadtree method, then combines the improved no search method with improved quadtree method to greatly speed up the fractal encoding process and hold high reconstruction fidelity. Some other time-consuming parts of fractal coding are redesigned and accelerated with new techniques. Experiments on standard images show that the proposed scheme can reduce encoding time greatly with only a little loss of approximation quality.

An on-line unsupervised learning mechanism is proposed for unlabeled data which is polluted by noises. By using a similarity threshold and local error based insertion criterion, the system is able to grow incrementally and to accommodate input patterns of on-line non-stationary data distribution. The definition of a utility parameter - "error-radius" enables this system to learn the number of nodes needed to solve a current task. The usage of a new technique for removing nodes in low probability density regions can separate the clusters with low-density overlaps and dynamically eliminate the noise in the input data. The design of two-layer neural network makes it possible for this system to represent the topological structure of unsupervised on-line data, report the reasonable number of clusters and give typical prototype patterns of every cluster without any priori conditions such as suitable number of nodes or a good initial codebook.

This paper introduces a kernel logit approach for face and non-face classification. The approach is based on the combined use of the multinomial logit model (MLM) and "kernel feature compound vectors." The MLM is one of the neural network models for multiclass pattern classification, and is supposed to be equal or better in classification performance than linear classification methods. The "kernel feature compound vectors" are compound feature vectors of geometric image features and Kernel features. Evaluation and comparison experiments were conducted by using face and non,face images (Face: training 100, cross-validation 300, test 325, Non-face : training 200, cross-validation 1000, test 1000) gathered from the available face databases and others. The experimental result obtained by the proposed method was better than the results obtained by the Support Vector Machines (SVM) and the Kernel Fisher Discriminant Analysis (KFDA).

This paper describes briefly an algorithm both for real-time facial image recognition in unconstrained indoor environments and a human like agent image synthesis which responds to the recognition results. Our simple but effective image recognition algorithm can recognize facial images from moving images, choose one primal face and track the primal face in real-time. Therefore, the agent can gaze/glance/etc., at the primal face (the interacting user) during the process. It means a "facial contact" or an "eye contact" can be achieved between the user and the synthesized moving agent, which is very important for friendly human interfaces.

Describes a parallel computing system and a software algorithm for realtime interaction between a human user and a synthesized moving humanlike image or agent. The realistic human-like agent appearing on a display can recognize the palm position and finger motion (finger sign) of the human user, then tracks and gazes at the hand position and changes his/her facial expression in response to the recognized finger sign in realtime. This kind of interactive agent is expected to play an important role in advanced human interfaces. To realize realtime image recognition and image synthesis, a parallel image processing and synthesis system named 'TN-VIT' has been developed.>

An automatic organisms recognition and identification method for a micromanipulator system, using image processing based on an expert system, is described. The proposed method is based on the organism image segmentation method (OISM), which takes advantage of characteristic segment features that are independent of individual size and length. The complicated shapes of the organisms are divided into basic shape segments such as lines, circles, ovals, etc. The relation between segments are analyzed and described in the database of the expert system automatically. Organisms are then expressed simply by a set of segments, so that their individual differences are avoided. Tracking the movement of an organism by its image is also shown. The effectiveness of this method is confirmed by experimental results.>