Mariusz Rybnik

K3M: A universal algorithm for image skeletonization and a review of thinning techniques

K3M: A universal algorithm for image skeletonization and a review of thinning techniques This paper aims at three aspects closely related to each other: first, it presents the state of the art in the area of thinning methodologies, by giving descriptions of general ideas of the most significant algorithms with a comparison between them. Secondly, it proposes a new thinning algorithm that presents interesting properties in terms of processing quality and algorithm clarity, enriched with examples. Thirdly, the work considers parallelization issues for intrinsically sequential algorithms of thinning. The main advantage of the suggested algorithm is its universality, which makes it useful and versatile for a variety of applications.

User Authentication with Keystroke Dynamics Using Fixed Text

User authentication is an important problem for any restricted access computer system, especially with remote access. Constant spread of Internet makes efficient remote authentication a pressing issue while the distributed nature of interaction introduces additional difficulties. Authentication methods based on most biometrics techniques require dedicated hardware that is unhandy for remote applications. In contrast keystroke dynamics requires only limited dedicated software and does not require any dedicated hardware. The approach we propose is aimed at efficient user authentication with keystroke dynamics using short fixed text, that may for example occur simultaneously to logging process. We have tested the suggested approach on a large group of individuals, with data gathered over Internet using browser-based WWW application. The obtained results are promising and encourage further development in this area.

An Exploration of Keystroke Dynamics Authentication Using Non-fixed Text of Various Length

User authentication is an important problem for any restricted access computer system, especially with remote access. Constant spread of Internet makes efficient remote authentication a pressing issue with remote interaction providing additional difficulties. Authentication methods based on most biometrics techniques mostly require dedicated hardware that is unhandy for remote applications. Keystroke dynamics however requires uniquely dedicated software and no indispensable dedicated hardware. The approach proposed by the authors in this paper is aimed at efficient user authentication with keystroke dynamics using non-fixed text of various size. We have tested the approach on a small group of individuals, with data gathered in various ways: over Internet using browser-based WWW application and on local machines using dedicated applications. The obtained results cans be used as indication for future keystrokes database creation.

Querying in spaces of music information

This study is focused on querying accomplished on structured spaces of information. Querying is understood in terms of mining structures of information and of knowledge understanding. We consider information as a subject of descriptions expressed in some language. Information is hidden behind such descriptions. Operations done on structured spaces of information are performed on language constructions describing such structures. However, automatic operations not always can be performed directly on language constructions. In such cases it is necessary to expand performance to the space of information. The study concerns paginated (i.e. printed and handwritten) music notation. It is shown that querying in the space of music information requires syntactic structuring as well as its expansion to semantic analysis. It is worth underlining that data understanding requires analysis of uncertainty: analyzed data are usually incomplete, uncertain and with some incorrectness. Such imperfectness of information is hidden under the level of syntax and semantics. Due to limitation of the paper this problem is not studied.

User Authentication for Mobile Devices

The paper is intended as a short review of user authentication problem for mobile devices. The emphasis is put on smartphones and tablets, that nowadays are very similar to miniaturized personal computers with much more sensors of various origin. The sensors are described with remarks on their usefulness for user authentication. Deficiencies of traditional user authentication methods based on knowledge are pointed out and the need for new – more secure but also comfortable – user authentication mechanisms is reasoned. Preliminary user authentication systems employing biometric features are discussed and hence the generally unused potential of biometrics for mobile devices is demonstrated.

Signature System Based on Extended Shape Context Descriptors

In this paper we present a system for handwritten signature identification and verification based on Shape Context Descriptors. The input to the system are binarized images of handwritten signatures. At preprocessing stage each signature image is thinned using KMM algorithm in order to obtained 1-pixel wide skeleton. The feature vector is built from Shape Context Descriptors computed for selected points on skeletonized signature line. The process of identification and verification is based on the distance measure that uses Shape Context Descriptors. The experiments were conducted on signature images from GPDS database available for non-commercial research.

Advanced searching in spaces of music information

This study continues construction of specialized grammars covering music information. The proposed grammar is searching-oriented and therefore largely simplified, in order to create easily-searchable structure. Searching is seen here as a particular querying operation in spaces of music information. Searching is discussed for patterns melodically and rhythmically transformed - with transformations in pitch and time dimensions typical for music works. Three operators are proposed to provide convenient meta-data for searching. Searching may be performed in both structured and unstructured musical pieces, regarding voice selection. Proposed methodology may serve for searching of transformed and non-transformed motives, searching of inspirations, comparative analysis of musical pieces, analysis of melodic and rhythmical sequences, harmonic analysis and structure discovery.

FE8R - A Universal Method for Face Expression Recognition

This paper proposes a new method for recognition of face expressions, called FE8R. We studied 6 standard expressions: anger, disgust, fear, happiness, sadness, surprise, and additional two: cry and natural. For experimental evaluation samples from MUG Facial Expression Database and color FERET Database were taken, with addition of cry expression. The proposed method is based on the extraction of characteristic objects from images by gradient transformation depending on the coordinates of the minimum and maximum points in each object on the face area. The gradient is ranked in \([-15,+35]\) degrees. Essential objects are studied in two ways: the first way incorporates slant tracking, the second is based on feature encoding using BPCC algorithm with classification by Backpropagation Artificial Neural Networks. The achieved classification rates have reached 95 %. The second method is proved to be fast and producing satisfactory results, as compared to other approaches.

Dealing with contaminated datasets: An approach to classifier training

The paper presents a novel approach to classification reinforced with rejection mechanism. The method is based on a two-tier set of classifiers. First layer classifies elements, second layer separates native elements from foreign ones in each distinguished class. The key novelty presented here is rejection mechanism training scheme according to the philosophy “one-against-all-other-classes”. Proposed method was tested in an empirical study of handwritten digits recognition.

A harmonization model with partial fuzzy knowledge

The paper presents a knowledge-driven harmonization model for tonal music partially using fuzzy knowledge. Automatic harmonization could be seen as a complicated classification problem: an algorithm processes a music notation document and generates class labels (harmonic functions). The proposed model could be seen as an Expert System, based largely on the music theory with partial fuzzy knowledge. It emphasizes universality - a possibility of obtaining varied but controllable harmonies. It is directly tunable by changing the internal parameters of harmonization mechanisms, as well as an importance weight corresponding to each mechanism. The proposed harmonization model allows also for relatively easy interpretation of internal knowledge.