2004/3
Browse
Recent Submissions
- ItemMAP Based Speaker Adaptation in Very Large Vocabulary Speech Recognition of Czech(Společnost pro radioelektronické inženýrství, 2004-09) Cerva, Petr; Nouza, JanThe paper deals with the problem of efficient adaptation of speech recognition systems to individual users. The goal is to achieve better performance in specific applications where one known speaker is expected. In our approach we adopt the MAP (Maximum A Posteriori) method for this purpose. The MAP based formulae for the adaptation of the HMM (Hidden Markov Model) parameters are described. Several alternative versions of this method have been implemented and experimentally verified in two areas, first in the isolated-word recognition (IWR) task and later also in the large vocabulary continuous speech recognition (LVCSR) system, both developed for the Czech language. The results show that the word error rate (WER) can be reduced by more than 20% for a speaker who provides tens of words (in case of IWR) or tens of sentences (in case of LVCSR) for the adaptation. Recently, we have used the described methods in the design of two practical applications: voice dictation to a PC and automatic transcription of radio and TV news.
- ItemAnalysis of Homogeneous Coplanar Strip Line(Společnost pro radioelektronické inženýrství, 2004-09) Sadek, Vaclav; Svacina, JiriThe goal of this work is to introduce a new, maybe complicated but in the final result mathematically simplest model of the coplanar strip line (CPS). In contrast to the usual method based on elliptical integrals the simplest circular inversion is applied. The main advantage is that our solution is mathematically less complicated but its accuracy is a little bit lower. The maximal error of the model described lies within the restricted interval between -3% and 3%. Nevertheless the final formula is useful for the practical engineering application.
- ItemBAMUD Features Demonstration by System View(Společnost pro radioelektronické inženýrství, 2004-09) Longauer, Leos; Marchevsky, Stanislav; Kocur, DusanDirect-sequence code-division multiple access (DS-CDMA) is a frequently used wireless technology in DS-CDMA communications. The conventional DS-CDMA detector follows a single-user detection strategy in which each user is detected separately without regard for the other users. The better strategy is multi-user detection (MUD), where information about multiple users is used to improve detection of each individual user. This paper presents an adaptive multi-user detector converging (for any initialization) to the minimum mean square error (MMSE) detector without requiring training sequences. This blind multi-user detector (BAMUD) requires no more knowledge than does the conventional single-user detector. The structure of adaptive blind detector is simulated by the system design tool SystemView. The aim focus is to verify theoretical knowledge of BAMUD structure using hardware-oriented PC-based model in SystemView.
- ItemBehind the Structure of Video VQ-Coder(Společnost pro radioelektronické inženýrství, 2004-09) Hanzlik, Pavel; Pata, PetrThis paper introduces an implementation and structure of a video codec (coder-decoder) using Vector Quantization (VQ) in the program Matlab. It also aims on the VQ weak and strong features and it describes an experiment with turning the advantages into concrete profit in video data compression. We would like to aim also on usage in video MPEG 4 (Motion Picture Experts Group) standard or in compressing the scientific (high resolution) images. In conclusion we are comparing our approach with other image compression methods with objective and subjective criteria of image quality perception.
- ItemCorrelation Analysis in a Pulse Wave Velocity Evaluation(Společnost pro radioelektronické inženýrství, 2004-09) Jelinek, Martin; Dobes, Josef; Pousek, LubomirIn this paper, methods for a time delay evaluation of phonocardiographic (PCG) signals are presented to estimate a pulse wave velocity (PWV) in a cardiovascular system of a human body, especially in arterial segments of an arterial tree selected. A measuring method used for the pulse wave registration is fully non-invasive. Electronic phonendoscopes pressure/acoustic converters were used as signal transducers. The PWV estimation was carried out using correlation analysis of PCG signals, square of raw PCG signals and the first derivations of PCG signals. Signal processing, i.e. filtration, standardization, etc. was implemented in a Matlab environment using created application. A set of subjects examined in this experiment consists of five young healthy volunteers.