Faculty of Electrical Engineering

The threshold sensitivity of the leg-vibration receptors of the green lacewing, Chtysoperla carnea, was investigated electro-physiologically. The legs were stimulated by sinusoidal vibrations. Summed responses from the leg nerve fibres of different legs do not differ much. In each leg there are two scolopidial organs which respond to vibration. The most sensitivethe subgenual organ-shows maximum sensitivity in the frequency range 1.5-2 kHz and a displacement threshold near 0.2 nm, corresponding to an acceleration of about 0.03 ms-'. The other is the femoral chordotonal organ with maximum sensitivity at about 1 kHz and the threshold between 0.1 and 1 msP2. After ablation of both scolopidial organs, only responses of non-scolopidial organs persist, very probably campaniform sensilla.

The onion plant (Allium cepa L.) is a suitable indicator plant for the determination of potential genotoxic agents in the samples taken from the environment. The genotoxic level of the agent under study is reflected by structural changes of the chromosomes and their changed numbers. The chromosomes under study are taken from the meristem cells of the young growing onion roots. A healthy normal onion cell has 16 (2n = 16) chromosomes. They are relatively large and so very appropriate for the detection of morphological changes. Prior to the chromosome study the root tip cells were immersed in a 0.1% aquatic solution of colchicine which stopped the mitotic cycle continuing beyond metaphase. The changes in morphology varied from a single distortion of a single chromosome up to several morphological changes observed on many chromosomes. We identified 15 categories of morphological aberrations which are classified into three groups: chromatid damage (CtD), centromere damage (CmD) and chromosome damage (CsD). CtD includes: single break chromatid, double break chromatid, isochromatid break, multiple break chromatid, gap chromatid, centric ring chromatid, acentric ring chromatid and triradial chromosomes. CmD includes: break centromere, gap centromere, single break centromere, double break centromere and multiple break centromere. CsD includes: ring chromosomes and dicentric chromosomes. Sometimes also the chromosome number changed which occurred as aneuploidy with monosomy 2n = 15 (2n =16-1) and euploidyincreased number of the basic chromosome number (2n = 6x-8x). We identified also the translocation: t(3p-; 5p+).

Web3D or virtual reality on the web used interactive, immersive three-dimensional worlds, not to be mistaken with video clips. We write here about our experiences and suggestions on this attractive technique, which we use and develop for more than ten years. Our field of work is natural science e.g. biology.

The Clark Aldrich's book: "Simulations and the Future of Learning : An Innovative (and Perhaps Revolutionary) Approach to e-Learning" is about new educational ways that are tightly connected to computer games and simulations. Clark Aldrich has created a detailed study how to design and realize an e-learning simulation that is both educative and similar to a modern computer game.

An edit-distance model that can be used for the approximate matching of contiguous and noncontiguous timed strings is presented. The model extends the concept of the weighted string-edit distance by introducing timed edit operations and by making the edit costs time dependent. Special attention is paid to the timed null symbols that are associated with the timed insertions and deletions. The usefulness of the presented model is demonstrated on the classification of phone-recognition errors using the TIMIT speech database.

Variations of the basic string-alignment algorithm are commonly used for the detection and classification of speech-recognition errors. In this procedure, reference and system-output hypothesis speech transcriptions are first aligned using the string-alignment algorithm that is based on primitive edit operations. The edit operations needed to transform one transcription into the other are then tallied as speech-recognition errors. The algorithms normally detect approximately the same total number of errors; however, they can produce different error classifications. This paper investigates these differences and several criterion functions that can be used for the comparison and evaluation of the algorithms for the detection and classification of speech-recognition errors. The proposed criterion functions were used for the experimental evaluation of the standard algorithms that are implemented as part of the CUED HTK and the NIST SCTK, and were used for the detection and classification of phone-recognition errors from the TIMIT speech database.

The workshop on Restrictions in the Implementation of EU Data Protection Directive for Public Interest, Security and Defence, was the second of the two problem-solving workshops to be held in the HIDE project. This one-and-a-half-day workshop took place at Grand Hotel Union – Hotel and Conference Centre, Miklošičeva 1, Ljubljana, Slovenia on 17-18 September 2009. The workshop was organized by the Laboratory of Artificial Perception, Systems and Cybernetics, Faculty of Electrical Engineering, University of Ljubljana, in cooperation with the Centre for Science, Society and Citizenship (CSSC), Italy. This paper summarizes the results of the workshop.

Blind and visually-impaired people face many problems in interacting with information retrieval systems. State-of-the-art spoken language technology offers potential to overcome many of them. In the mid-nineties our research group decided to develop an information retrieval system suitable for Slovene-speaking blind and
visually-impaired people. A voice-driven text-to-speech dialogue system was developed for reading Slovenian texts obtained from the Electronic Information System of the Association of Slovenian Blind and Visually Impaired Persons Societies. The evolution of the system is presented. The early version of the system was designed to deal explicitly with the Electronic Information System where the available text corpora are stored in a plain text file format without any, or with just some, basic non-standard tagging. Further improvements to the system became possible with the decision to transfer the available corpora to the new web portal, exclusively dedicated to blind and visually impaired users. The text files were reformatted into common HTML/XML pages, which comply with the basic recommendations set by the Web Access Initiative. In the latest version of the system all the modules of the early version are being integrated into the user interface, which has some basic web-browsing functionalities and a text-to-speech screen-reader function controlled by the mouse as well.

The paper first presents a large-vocabulary automatic speech recognition system that is being developed for the Slovenian language. The concept of a single-pass token-passing algorithm for the fast speech decoding that can be used with the designed multi-level system structure is discussed. From the algorithmic point of view, the main component of the system is a finite-state pronunciation lexicon model. This component has crucial impact on the overall performance of the system and we developed a sequential minimization algorithm that very efficiently reduces the size and algorithmic complexity of the lexicon model. Our finite-state lexicon model is represented as a state-emitting finite-state transducer. The presented experiments show that the sequential minimization algorithm easily outperforms (up to 60%) the conventional algorithms that were developed for the static global optimization of the transition emitting finite-state transducers. These algorithms are delivered as part of the AT&T FSM libraryTM and the OpenFST library.

The paper presents our experiences with the phone transition acoustical models. The phone transition models were compared to the traditional context dependent phone models. We put special attention on the speech signal segmentation analysis to provide a better insight into certain segmentation effects when using the different acoustical models. Experiments with the HMM-based models were performed using the HTK toolkit, which was extended to allow proper state parameter tying for the phone transition models. All the model parameters were estimated on the GOPOLIS speech database. The annotation
confusions concerning two-phone speech units are discussed. Finally, the overall word recognition score is
presented. The better score was achieved using the diphone models even when comparing them to the triphone models.

This paper presents the results of our efforts to obtain the minimum possible finite-state representation of a pronunciation dictionary. Finite-state transducers are widely used to encode word pronunciations and our experiments revealed that the conventional redundancy-reduction algorithms developed within this framework yield suboptimal solutions. We found that the incremental construction and redundancy reduction of acyclic finite-state transducers creates considerably smaller models (up to 60\%) than the conventional, non-incremental (batch) algorithms implemented in the OpenFST toolkit.

In this article we present an efficient approach to modeling the acoustic features for the tasks of recognizing various paralinguistic phenomena. Instead of the standard scheme of adapting the Universal Background Model (UBM), represented by the Gaussian Mixture Model (GMM), normally used to model the frame-level acoustic features, we propose to represent the UBM by building a monophone-based Hidden Markov Model (HMM). We present two approaches: transforming the monophone-based segmented HMM-UBM to a GMM-UBM and proceeding with the standard adaptation scheme, or to perform the adaptation directly on the HMM-UBM. Both approaches give superior results than the standard adaptation scheme (GMM-UBM) in both the emotion recognition task and the alcohol detection task. Furthermore, with the proposed method we were able to achieve better results than the current state-of-the-art systems in both tasks.