Human – Computer Systems Interaction: Backgrounds and Applications 2

It is demonstrated that spectral characteristics of steady-state visual evoked potentials (SSVEPs) in a brain-computer interface (SSVEP-based BCI) depend significantly on the stimulus parameters, such as color and frequency of its flashing light. We postulate these dependencies can be used to improve the BCI performance – by proper design, configuration and adjustment of the visual stimulator. Preliminary results of conducted experiments show also that SSVEP characteristics are strongly affected by subjects biodiversity.

The main goal of this chapter is to propose a guideline for human-robot systems focused on ergonomic haptic interaction. With this aim, this model presents several main parts: a set of heuristic indicators in order to identify the attributes of the haptic interaction, the relationship between indicators, the human task and the haptic interface requirements and finally an experimental task procedure and a qualitative performance evaluation metrics in the use of haptic interfaces. The final goal of this work is the study of possible applications of haptics in regular laboratory conditions, in order to improve the analysis, design and evaluation of human task over haptic interfaces in telerobotic applications.

Toyota has been developing industrial robots since the 1980s. In recent years, man-machine cooperative robots that assist people’s skills have gradually been put into practical use. Now, Toyota is further evolving a number of robot technologies born at production sites to develop Partner Robots that work in harmony with people. Toyota is specifically considering four development areas: (1) manufacturing support, (2) short-distance personal mobility, (3) nursing and healthcare support, and (4) support for work around the home. Some scenes of Partner Robot operation in each of these four fields are mentioned, as well as the essential technologies embedded in Toyota Partner Robots, are then introduced in this paper. Lastly, issues that need to be resolved to bring these robots into truly useful and practical in our society are described.

An intelligent robot has been developed which understands the goal a user wants to be met, recognizes its environment, develops strategies to achieve the goal and operates autonomously. By means of a speech recognition sensor the robot listens to the command spoken by a user and derives the goal, i.e. the task the user wants the robot to perform such as to bring a specific object. Next, the robot uses its smart camera and other sensors to scan the environment and to search for the demanded object. After it has found and identified the object, it grabs it and brings it to the user. Additionally, a method for generating a will is proposed. This enables the robot to operate optimally even under conflicting requirements.

A talking and singing robot which adaptively learns the vocalization skill by an auditory feedback learning is being developed. The fundamental frequency and the spectrum envelope determine the principal characteristics of a sound. The former is the characteristics of a source sound generated by a vibrating object, and the latter is operated by the work of the resonance effects. In vocalization, the vibration of vocal cords generates a source sound, and then the sound wave is led to a vocal tract, which works as a resonance filter to determine the spectrum envelope. The paper describes the construction of vocal cords and a vocal tract for the realization of a talking and singing robot, together with the control algorithm for the acquisition of singing performance by mimicking human vocalization and singing voices. Generated voices were evaluated by listening experiments.

Recent research in orthopedic surgeries indicates that computer-assisted robotic systems have shown that robots can improve the precision and accuracy of the surgery which in turn leads to better long-term outcomes. Increasing demand on minimal invasive bone cutting operation has been encouraging surgical robot developments in orthopedics. In this work, an orthopedic robotic system called OrthoRoby and an intelligent control architecture that will be used in bone cutting operations are developed. Experiments are performed to demonstrate the performance of the intelligent control architecture.

In this study, a bilateral control system for two-degree-of-freedom (two-DOF) remote robots that are capable of grasping and manipulating motion is considered. The purpose of this research is to achieve force-sensor-less bilateral control with thrust wires for two-DOF systems with small operational forces. Small operational forces in remote robot systems are suitable for several applications. In conventional research, methods for reducing the operational forces in one-DOF systems with thrust wires were proposed. In this study, this method is applied to a two-DOF system. Furthermore, a method for further reduction of operational forces is proposed in which force transforms of both local and modal space are implemented. By considering modal space in the two-DOF system, the operational forces can be reduced further. The validity of the proposed method is confirmed using experiments.

Computational intelligence is finding more and more applications in computer aided diagnostics, helping doctors to process large quantities of various medical data [Buronni et al. 2004]. In dermatology it is extremely difficult to perform automatic diagnostic differentiation of malignant melanoma based only on dermatoscopic images. Applying artificial intelligence algorithms to explore and search large database of dermatoscopic images allow doctors to semantically filter out image with specified characteristics. This paper presents an semantic approach for characteristic objects classification found in image database of pigment skin lesions, based on radial basis function kernel for artificial neural networks. Presented approach is divided into few parts: JSEG image segmentation [Deng et al. 2001], feature extraction and classification. Prepared features vector consist of color models parts. For classification Artificial Neural Networks and Support Vector Machines are used and their performance is evaluated and compared. Success rates in both cases are greater than 90%.

This paper presents a method for automatic identification of asymmetry in digital images containing melanocytic skin lesion. Our method is a part of the new system for classifying skin lesion using Stolz strategy, based on the ABCD rule.

We have described the advantages of the multispectral imaging in different applications, outlined its challenges and analyzed approaches to multispectral objects detection. We have proposed, built and deployed multispectral acquiring device with liquid crystal tunnable filter for the endoscopy diagnosis. We have applied the device in the photodynamic diagnosis for cancer detection based on the spectral pixel signatures and supervised machine learning techniques. We have evaluated introductory step in the spectrum matching approach - spectrum estimation, by linear transformations of the pixels spectral signatures. We have used image dataset acquired of the GretagMacbeth colorchecker. We have examined visualization methods of multispectral image, helpful for a gastroscopy and co lonoscopy diagnostician. We have tried to calculate colour image by the linear regression methods.

Learning models related to brain stroke data of 162 anonymous patients were generated in the form of sets of production rules using for selecting of descriptive attributes Glasgow Outcome Scale (

GOS

) and Modified Rankin Scale (

mRS

). The developed set of rules were then optimized leading to high accuracy rules, having distinctly decreased number of logic condition.

The aim of this study is to present the application of the eye-tracking technology to the research on autistic spectrum disorders (ASD) with a special interest on language impairments and text comprehension and production deficits. We discuss data and results obtained from a single case study research regarding an adult autistic male with a hyperlexia profile. Our results support the usage of eye-tracking technology in research and diagnostic contexts that make difficult an intrusive human-machine interaction.

The paper gives an overview of research devoted to developing a semi-automatic methodology of building a semantic model of medical diagnostic knowledge. The methodology is based on natural language processing methods which are applied to analyze medical texts. As a result of the process, the semantic model of symptoms is generated. This model is a foundation for building a model of diagnostic technologies. The described methodology and the resulting model are developed specifically for the Polish language.

Our research concerns psychometric data coming from the Minnesota Multiphasic Personality Inventory (MMPI) test. MMPI is used to count the personality-psychometric dimensions which help specialists in diagnosis of mental diseases. In this paper, we present a part of the Copernicus system – the tool for computer-aided diagnosis of mental diseases based on personality inventories. This part is devoted to the rule-based analysis of the MMPI data expressed in the form of the so-called profiles. The paper characterizes the knowledge base embodied in Copernicus which can be used for the rule-based analysis of the patients’ MMPI data as well as the functionality of the designed tool.

We present an interpolation algorithm for digital images based on the extended Edge-Directed Interpolation. The proposed algorithm utilises localisation and orientation of edges extracted with wavelet edge extractors. The algorithm chooses an interpolation method of a given point of an image, depending on localisation with respect to extracted edges. We apply our method to grey scale digital images. In this paper we propose also some modifications to the method. We present the results obtained with our algorithm, in comparison to a few popular interpolation algorithms and we propose a plan of further algorithm extensions.

This chapter presents aspects concerning the design of model-based fuzzy control solutions dedicated to the longitudinal slip control of an Antilock Braking System laboratory equipment. Continuous-time and discrete-time Takagi-Sugeno (T-S) fuzzy models of the controlled process are first derived on the basis of the modal equivalence principle. The consequents of the T-S models of the T-S fuzzy controllers are local state feedback controllers which are solutions to several linear quadratic regulator (LQR) problems and the parallel distributed compensation is next applied. Linear matrix inequalities are solved to guarantee the global stability of the discrete-time fuzzy control systems and to give the optimal state feedback gain matrices of the LQR problems. A set of real-time experimental results is included to validate the new fuzzy control solutions.

In this article the e-learning platforms applied in higher education are being looked at. The comparative analysis of certain platforms in respect of their use in a teaching process was made. The analysis was created based on criteria, chosen for the easiest management of the groups of students, easier communication between lecturers and student, as well as for the easy forming of archives and transforming prepared materials in the SCORM/AICC standards. The aim of the article is to facilitate the school management’s choice of right systems. This is depending on the needs, as well as pointing out the possibilities of the e-learning platforms. The marking scale presented in the comparison of the platforms clearly confirms that the best platforms are ZSZN. Three systems: BlackBoard, Moodle and ILIAS, are ranked after ZSZN. Depending on the range of the distance learning systems used, the selection among these three platforms depends on the financial capability of a given higher education institution. BlackBoard is a system that provides an exhaustive offer in both asynchronous and synchronous e-learning. ILIAS and Moodle are open systems that support many plug-ins relating to functions that we may need. The paper offers a comparison of the functionality of the most popular e-learning systems. The comparative analysis also looks at the ZSZN system that has been designed mainly with universities of technology in mind. On the basis of the comparative analysis results, the university management will be able to choose an optimum e-learning system that is best adapted to cater to both their needs and financial capabilities.

The chapter offers some points of view concerning the education in control engineering on the basis of a Magnetic Levitation System with Two Electromagnets (MLS2EM) laboratory equipment. The syllabus of control engineering includes the treatment of the following issues: plant analysis and modeling, development of low-cost control solutions and algorithms applied on real-time laboratory experiments, and the assessment used in the syllabus of control engineering courses. The presentation is application-oriented focusing on the ML2EM laboratory equipment, and the low-cost control solutions presented here deal with state feedback control, proportional-integral-derivative control and model predictive control. Our syllabus structure is attractive as it allows the proper assessment of students’ knowledge. The real-time experiments highlighted in this chapter accompany the laboratory-based education in control engineering and they ensure the improved understanding of the theoretical aspects taught at the lectures.

Creative destruction–the creation of new industries and the destruction of old industries–is a very abstract concept. Those teaching entrepreneurship, where creative destruction is a central feature, often struggle to communicate the dynamism of industry evolution where industry disruption can yield innovation, entrepreneurial opportunities and new wealth. This paper examines the application of human-computer interaction (HCI) and specifically information visualization to entrepreneurship education, a specialized area of business education. We create and evaluate different 2-D and 3-D visualizations of industry evolution in the Popular Music Industry between 1951 and 2008 to determine which visualizations correspond to superior comprehension of creative destruction. Particularly, our challenge was to represent the emergence of 13 major markets and 193 submarkets in the context of six decades of music industry evolution and disruption. The results suggest information visualization is a resource for entrepreneurship education and that significant improvements can be made over current idiosyncratic methods of representing industry evolution.

The following Chapter presents a new approach to effective learning by employing a biofeedback method based on hemispherical synchronization. The application proposed uses a wireless EEG (

electroencephalography

) recording system of the user’s brain waves and powerful signal processing and classification to produce a reliable feedback. Alpha and beta brain rhythms are analyzed by applying DWT (

Discrete Wavelet Transform

) and by calculating the statistics for each analyzed window. EOG (

electrooculogram

) artifacts are eliminated from the signal through adaptive filtration in the time-frequency domain. Three different learning methods are implemented in the proposed application: mind map, flash cards and non-linear notes. Several tests are performed with the users. Based on the brain feedback information and the user’s learning profile test results, an optimized learning method is chosen for an individual user. Information about hemispherical synchronization provides vital information for system adjustments. The results obtained show a difference between traditional learning and one using a feedback loop, indicating that synchronized hemispheres improve learning abilities. In conclusion the critical evaluation of the method is given.

In this paper, a comparison between different fuzzy and neural systems is presented. Instead of using traditional membership functions, such as triangular, trapezoidal and Gaussian, in fuzzy systems, the monotonic pair-wire or sigmoidal activation function is used for each neuron. Based on the concept of area selection, the neural systems can be designed to implement the identical properties like fuzzy systems have. All parameters of the proposed neural architecture are directly obtained from the specified design and no training process is required. Comparing with traditional neuro-fuzzy systems, the proposed neural architecture is more flexible and simplifies the design process by removing division/normalization units.

The chapter presents hardware implementation of the model of commonsense reasoning system based on a new formalism Fuzzy Default Logic (FDL). It briefly recalls main definitions and algorithms of FDL technique in a form of software oriented procedures. Basic building blocks used for implementation are presented. Then the software algorithms of the FDL model are transformed into hardware blocks. The entire hardware structure has been implemented in FPGA Xilinx Virtex5 device. The obtained results, examples of experiments and applications in system verification platform are discussed and further research tasks formulated.

The paper includes a description of a novel method for reducing the size of a training set in order to reduce memory requirements and classification complexity. Our method allows the condensing of the training set in a way that it is both training set consistent (classifies all training data points correctly) and decision-boundary consistent (the decision boundary does not changes after applying our method) for NN classifiers. Furthermore, the algorithm described here allows the utilization of a parallel computing paradigm in order to increase performance.

While the original design of wikis was mainly focused on a completely open free-form text model, semantic wikis have since moved towards a more structured model for editing: users are driven to create ontological data in addition to text by using ad-hoc editing interfaces. This paper introduces OWiki, a framework for creating ontological content within

not-natively-semantic

wikis. Ontology-driven forms and templates are the key concepts of the system, that allows even inexpert users to create consistent semantic data with little effort. Multiple and very different instances of OWiki are presented here. The expressive power and flexibility of OWiki proved to be the right trade-off to deploy the authoring environments for such very different domains, ensuring at the same time editing freedom and semantic data consistency.

In this paper, a problem of MIMO object identification expressed mathematically in terms of fuzzy relational equations is considered. The identification problem consists of extraction of an unknown relational matrix, and also parameters of membership functions included in fuzzy knowledge base, which can be translated as a set of fuzzy IF-THEN rules. In fuzzy relational calculus this type of the problem relates to inverse problem and requires resolution for the composite fuzzy relational equations. The search for solution amounts to solving an optimization problem using a genetic algorithm. The resulting solution is linguistically interpreted as a set of possible rules bases. The approach proposed is illustrated by computer experiment and examples of diagnosis and prediction.

Protein structure similarity searching is a complex process, which is usually carried out through comparison of the given protein structure to a set of protein structures from a database. Since existing database management systems do not offer integrated exploration methods for querying protein structures, the structural similarity searching is usually performed by external tools. This often lengthens the processing time and requires additional processing steps, like adaptation of input and output data formats. In the paper, we present our extension to the SQL language, which allows to formulate queries against a database in order to find proteins having secondary structures similar to the structural pattern specified by a user. Presented query language is integrated with the relational database management system and it simplifies the manipulation of biological data.

In this paper we prove some properties of new kind of rules – called generalized rules.

The purpose of this paper is to describe a new methodology dedicated to the analysis of boolean recommenders. The aim of most recommender system is to suggest interesting items to a given user. The most common criteria utilized to evaluate a system are its statistical correctness and completeness. The two can be measured by accuracy and recall indices. In this paper we argue that technical performance is an important step in the process of recommender system’s evaluation. We focus on four real-life characteristics i.e. time required to build a model, memory consumption of the built model, expected latency of creating a recommendation for a random user and finally the time required to retrain the model with new ratings. We adapt a recently developed evaluation technique, which is based on an iterative generation of bipartite graphs. In this paper we concentrate on a case when preferences are boolean, which is opposite to value-based ratings.

Multivariate regression models are often used for the purpose of prognosis. Parameters of such models are estimated on the basis of learning sets, where feature vectors (independent variables) are combined with values of response (target) variable. The values of response variable can be determined only with some uncertainty in some important applications. For example, in survival analysis, the values of response variable is often censored and can be represented as intervals. The interval regression approach has been proposed for designing prognostic tools in circumstances of such type of uncertainty. The possibility of using the convex and piecewise linear (CPL) functions in designing linear prognostic models on the basis of interval learning sets is examined in the paper.

Second order algorithms are very efficient for neural network training because of their fast convergence. In traditional Implementations of second order algorithms [Hagan and Menhaj 1994], Jacobian matrix is calculated and stored, which may cause memory limitation problems when training large-sized patterns. In this paper, the proposed computation is introduced to solve the memory limitation problem in second order algorithms. The proposed method calculates gradient vector and Hessian matrix directly, without Jacobian matrix storage and multiplication. Memory cost for training is significantly reduced by replacing matrix operations with vector operations. At the same time, training speed is also improved due to the memory reduction. The proposed implementation of second order algorithms can be applied to train basically an unlimited number of patterns.

In this paper dynamic global models of input-output complex systems are discussed. Dynamic complex system which consists of two nonlinear discrete time sub-systems is considered. Multilayer neural networks in a dynamic structure are used as a global model. The global model is composed of two sub-models according to the complex system. A quality criterion of the global model contains coefficients which define the participation of sub-models in the global model. The main contribution of this work is the influence study on the global model quality of these coefficients. That influence is examined for different back propagation learning algorithms for complex neural networks.