Dr. Khalid Saeed, DSc, PhD, MSc, BSc Engg.
Professor of Computer Science at AGH
Faculty of Physics and Applied Computer Science,
AGH University of Science and Technology
DVP term expires December 2013
Khalid Saeed received the BSc Degree in Electrical and Electronics Engineering in 1976 from Baghdad University in 1976, the MSc and PhD Degrees from Wroclaw University of Technology, in Poland in 1978 and 1981, respectively. He received his DSc Degree (Habilitation) in Computer Science from Polish Academy of Sciences in Warsaw in 2007. He is a
Professor of Computer Science with AGH University of Science and Technology in Poland. He has published more than 120 publications – 17 edited books and 7 text and reference books. He supervised about 30 BSc projects, 90 MSc and 9 PhD theses. His areas of interest are Image Analysis and Processing, Biometrics and Computer Information Systems.
Toeplitz Matrices in Biometric Recognition Systems
The talk will be on a recently worked out mathematical model for Image Annotation. It discusses the history of Biometrics, Spoofing and Forgery, Anti-Spoofing models and Multi-model Biometric systems. The author will also show the role of Toeplitz model in data transformation from the huge number of characteristic points to an easy for implementation feature vector represented by a sequence of less number of components.
When Biometric Systems Fail to be Reliable in Security Systems.
In this talk, the author will try to answer questions concerning problems in our everyday contact with Biometric systems. A solution is proposed by introducing examples of hybrid biometric systems. The main mathematical tool is Toeplitz matrices and the based on them techniques. This has
proved to easily reduce the object-image feature data and help get simple mathematically reduced version of image description for object recognition. This is very important in the biometrics feature representation stages of multi-model systems in human verification for personal identification. The
mathematical model is explained briefly and given in this work with some particular description of data reduction and transformation through an easy-to-implement algorithm. This theory may have its place among other known approaches and means of object data reduction and representation. It would hopefully lead to fast and practical methods of secure human identification, particularly in the case of multi-systems with integrated biometric methods where data size is huge.
Voice-Print and its Toeplitz-based Identifying Algorithms.
This talk will present new applications of Toeplitz matrix eigenvalues approach in image description, feature extraction and recognition. It discusses the possibility of treating the speech signal graphically in order to extract the essential image features as a basic step in successful data mining in the biometric techniques. The considered object here is the human-voice signal. The suggested frequency spectral estimation and Toeplitz-based approach, built on linear predictive coding principle,
has proved the possibility of selecting signal features from the power spectral plot and enter Toeplitz matrix in a similar manner to its other applications. Such applications on topics like images of written texts, signatures, palm-prints, face geometry or fingerprints have shown a success rate of about 98% in many cases. The extracted feature-carrying image comprises the elements of Toeplitz matrices to consecutively compute their minimal eigenvalues and introduce a set of feature vectors within a class of voices. The required computations are performed in MATLAB proving speech-signal image recognition in a simple and easy-to-use way as the system does not need to implement any special
hardware. This also implies that the suggested approach can be used in tandem with other biometric technologies in hybrid systems for multi-factor verification.
Toeplitz-Matrix Applications in Image Processing
In this talk new modifications and experiments for word recognition and classification are presented. The algorithm is based on recognizing the whole words without separating them into letters. The whole word is treated and analyzed as an image. The method is based on the
modification of a novel view-based word recognition algorithm – an approach that was successfully used by the authors’ in previous works. This method shows how to recognize words without segmentation. The top and bottom views of the word are analyzed in order to create the feature vector. Then the feature vector is processed by the aid of Toeplitz matrices. The obtained series of Toeplitz matrix minimal eigenvalues are used for classification. The results are promising.