Marcin Piekarczyk

Hierarchical Random Graph Model for Off-line Handwritten Signatures Recognition

The article presents the approach based on the usage of syntactic methods for the static analysis of handwritten signatures. The graph linguistic formalisms applied, such as the IE graph and ETPL(k) grammar, are characterised by considerable descriptive strength and a polynomial membership problem of the syntactic analysis. For the purposes of representing the analysed handwritten signatures, new hierarchical (two-layer) HIE graph structures based on IE graphs have been defined. The two-layer graph description makes it possible to take into consideration both local and global features of the signature. The usage of attributed graphs enables the storage of additional semantic information (in the form of a set of parameters) describing the properties of individual signature strokes. Information about the shapes of specimen signatures is presented in the form of a language based on random IE graphs and stochastic class ETPL(k) grammars, which makes it possible to take into consideration the natural variability of the shape of specimen signatures. The verification and recognition of a signature consists in analysing the affiliation of its graph description to the language describing the specimen database. Initial assessments display a precision of the method at a average level of about 17%.

On Using Palm and Finger Movements as a Gesture-Based Biometrics

In this paper we consider possibility of using palm movements as effective behavioral biometric modality. We propose to exploit the hand motion characteristics gathered from 3D sensor device as input data for identification system. Currently for various reasons people are concerned about directly touching the biometric scanners, therefore touch-less and non-invasive approach seems to be useful in practical applications. In the described scheme palm and fingertip positions are tracked in real-time, while the defined gesture is performed. The proposed gesture is composed of the spatially arranged well-known signs like letters and numbers. The applied matching algorithm utilizes a combination of DTW and DCT techniques for comparing data series. The experiments show promising results using the proposed method.

Matrix-Based Hierarchical Graph Matching in Off-Line Handwritten Signatures Recognition

In this paper, a graph-based off-line handwritten signature verification system is proposed. The system can automatically identify some global and local features which exist within different signatures of the same person. Based on these features it is possible to verify whether a signature is a forgery or not. The structural description in the form of hierarchical attributed random graph set is transformed into matrix-vector structures. These structures can be directly used as matching pattern when examined signature is analyzed. The proposed approach can be applied to off-line signature verification systems especially for kanji-like or ideogram-based structurally complex signatures.

Random Graph Languages for Distorted and Ambiguous Patterns: Single Layer Model

The work introduces a linguistic based model designed for distorted or ambiguous patterns where a graph based approach is used for structure representation. The knowledge about unevenness is usually created on the basis of finite number of patterns treated as positive samples of unknown language. The IE graphs are used as the base. Single pattern can be represented using deterministic IE graph. Subsequently, the collection of patterns, represented by deterministic graph is transformed into equivalent random graph language. Utilization of the grammatical inference mechanisms gives the possibility to perform this process in automatic way. Using the IE graphs and imposing some simple limitations on graph structures allows to obtain a polynomial complexity of knowledge inference. In the work it is described how to use the proposed model for collecting the knowledge in handwritten signatures recognition and analysis systems. Information about graphemes (solid fragment of handwritten signature) variability is stored in the form of random IE graphs and stochastic ETPL(k) graph grammars. Instead of an ordinary the IE graph, an attributed one is used in order to increase a descriptive power of the proposed schema. The parametrical data embedded in the graph carries some additional semantic information associated with the structure of pattern. The work presents discussion about inference scheme and computational complexity of the proposed linguistic representation scheme. Described methodology can be especially suited for creating the knowledge representation of the handwritten signatures, signs and ideograms (e.g. kanji) in offline recognition systems.

The Touchless Person Authentication Using Gesture-Types Emulation of Handwritten Signature Templates

The paper proposes a secure real-time user authentication system based on dynamic handwritten signature verification. We discuss the touch-less sensor-based authentication mechanism that relies on remote tracking palm-gestures imitating the handwritten signature pattern of examined person. The appropriate data is gathered from non-invasive sensor device in the form of time-ordered series related to spatial coordinates of the pen-like tool position and velocity. The proposed matching scheme exploits data series analysis in joint with feature-based classification. The discrete cosine transform is used to deal with instability of the genuine patterns. We also consider of using the local sensitivity hashing functions to obtain the better efficiency. The detailed analysis of the experimental results is included, too.

Human Actions Analysis: Templates Generation, Matching and Visualization Applied to Motion Capture of Highly-Skilled Karate Athletes

The aim of this paper is to propose and evaluate the novel method of template generation, matching, comparing and visualization applied to motion capture (kinematic) analysis. To evaluate our approach, we have used motion capture recordings (MoCap) of two highly-skilled black belt karate athletes consisting of 560 recordings of various karate techniques acquired with wearable sensors. We have evaluated the quality of generated templates; we have validated the matching algorithm that calculates similarities and differences between various MoCap data; and we have examined visualizations of important differences and similarities between MoCap data. We have concluded that our algorithms works the best when we are dealing with relatively short (2–4 s) actions that might be averaged and aligned with the dynamic time warping framework. In practice, the methodology is designed to optimize the performance of some full body techniques performed in various sport disciplines, for example combat sports and martial arts. We can also use this approach to generate templates or to compare the correct performance of techniques between various top sportsmen in order to generate a knowledge base of reference MoCap videos. The motion template generated by our method can be used for action recognition purposes. We have used the DTW classifier with angle-based features to classify various karate kicks. We have performed leave-one-out action recognition for the Shorin-ryu and Oyama karate master separately. In this case, 100% actions were correctly classified. In another experiment, we used templates generated from Oyama master recordings to classify Shorin-ryu master recordings and vice versa. In this experiment, the overall recognition rate was 94.2%, which is a very good result for this type of complex action.

Touch-Less Personal Verification Using Palm and Fingers Movements Tracking

In this paper the approach to personal authentication based on analysis of biomechanical characteristics related to palm movements is considered. The basic concept discussed in this research assumes that the hand motion dynamics, treated as a biometrics, can be a sufficient base for efficient user identification. As an input pattern for recognition system the natural finger-based gestures are investigated. The appropriate data is gathered from touch-less sensor device in the form of time-ordered data series related to spatial coordinates of fingertips positions and its velocities. The proposed matching scheme exploits data series analysis in joint with feature-based classification. The research is also focused on the analysis of such type of natural gestures which can be performed with as little awareness as possible. The possibility of using gestures performed in a high degree automatically and non-consciously can be considered as significant advantage in practical applications.

Usability of the Fuzzy Vault Scheme Applied to Predetermined Palm-Based Gestures as a Secure Behavioral Lock

Nowadays, the crypto-biometric schemes are commonly studied to deal with the key management problem existing in cryptographic systems. In the paper we consider a well-known fuzzy vault scheme working on non-standard biometric data associated with palm-oriented gestures. We discuss the usefulness of the touch-less palm tracking system focused on handling the biomechanical characteristics of ones hand to protect the cryptographic key or other secret data. As a behavioral template the fixed finger-based gesture is applied. In order to generate unordered sets for fuzzy vault we use the global approach where features are decoded from fingertip position and velocity. Both the raw and DCT-calculated time series are preprocessed to solve the instability and variability problem strictly related to behavioral character of the data. We provide the experimental results and discuss the security issues.

Simple System for Supporting Learning of Human Motion Capabilities

The paper presents an idea of the automatic system which is capable of improving the rehabilitation process of people suffering from diseases of locomotor system, and is able to enhance a person’s learning the motion skills. The simplest form of the system consists of a MEMS (Micro Electro-Mechanical Systems) sensor which detects movements of parts of the human body. The article describes possible methods of classification of the sensor signal. As a result of this process, the system sends to the training person a message concerning appropriate correction of his motion. The authors assert that the system, using even unsophisticated hardware (one sensor integrated with a microcontroller) and a simple minimum distance classifier, is able to improve the rehabilitation or the learning process. The article attempts to articulate the main conditions determining an increasing of the efficiency of the proposed idea. Namely: creation of a communication standard between sensors and the host computer used by a physician/teacher, and taking into account the whole multidimensional signal provided by MEMS (3-axis accelerometer, gyroscope and magnetometer).

Signal Recognition Based on Multidimensional Optimization of Distance Function in Medical Applications

The paper presents an idea of the method of creating the signal classifier which is based on the optimization of the metric (distance) function. The authors suggest that the proper choice of metric function parameters allows to adapt the whole classification operation to solve certain problems of the time-varied signal recognition, especially in medical applications. The main advantage of the described approach is a possibility to interpret the obtained solutions. This may enable to progress the doctor’s skills, as well as improve the automatic classification method. The paper presents a brief example of the method usage in a practical application. It deals with the classification of the signals obtained from MEMS (3-axis accelerometer) sensors during the Lachman knee test. The authors point to main conditions which determine an increase in the efficiency of the described approach. Particularly, they are involved in developing efficient optimization methods of discontinuous criterion functions and algorithms for detection the cohesive group of points that define the relevant signal regions.