Typically, click-evoked otoacoustic emissions (CEOAEs) are analyzed using fast Fourier transform which has greatly limited time-frequency (t-f) resolution. The purpose of the present study was to process CEOAEs recorded using linear and non-linear presentation modes by applying the Matching Pursuit (MP) method of adaptive approximation which provides a high t-f resolution. The data were recorded in 26 normal-hearing subjects using the ILO96 system with clicks presented at 78-82 dB pSPL. In the non-linear mode, a series of four clicks was delivered with three at the same level and polarity and the fourth three times greater in amplitude and inverted in polarity, whereas in the linear mode all stimuli were presented at the same level and polarity. The responses consisting of 512 data points were stored in two buffers. For each subject, 520 responses were recorded separately for off-line MP analyses. The MP method allowed decomposition of signals into waveforms of defined frequency, latency, time span, and amplitude and also identified patterns of resonance modes that were characteristic for CEOAEs recorded in each individual ear. The overall CEOAE levels were higher by 4 dB for the linear mode than those for the nonlinear method, in agreement with studies reported previously. In general, t-f properties of CEOAEs recorded with linear and non-linear protocols were similar with the exception of the 0-6-ms post-stimulus time window and the frequency range below 2.2 kHz. This part of the signal was contaminated by a stimulus artifact in the linear mode. The reproducibility factor grew faster with an increase of the number of averaged responses for the linear protocol than for the non-linear method. The results suggest that main differences between CEOAEs measured with the two methods are related to stimulus artifacts occurring at the beginning of the recording time-window in the linear mode.

A proper time-frequency transform technique suppresses the blurring and smearing effect of the time-varying Doppler shift on the target image. The conventional target imaging method uses the Fourier transform for extracting the Doppler shift from the received radar pulse. Since the Doppler shift is timevarying for rotating targets, the constructed images will be degraded. In this thesis, the Doppler shift information required for the Range-Doppler image of the target is extracted by using high resolution time-frequency transform techniques. The Wigner-Ville Distribution and the Adaptive Gabor Representation with the Coarse-to-Fine and the Matching Pursuit Search Algorithms are examined techniques for the target imaging system. The modified Matching Pursuit Algorithm, the Matching Pursuit with Reduced Dictionary is proposed which decreases the signal processing time required by the Adaptive Gabor Representation. The Hybrid Matching Pursuit Search Algorithm is also introduced in this thesis work and the Coarse-to-Fine Algorithm and the Matching Pursuit Algorithm are combined for obtaining better representation quality of a signal in the time-frequency domain. The stated techniques are applied on to the sample signals and compared with each other. The application of these techniques in the target imaging system is also performed for the simulated aircrafts.

Chen, Donoho a Saunders (1998) studují problematiku hledání řídké reprezentace vektorů (signálů) s použitím speciálních přeurčených systémů vektorů vyplňujících prostor signálu. Takovéto systémy (někdy jsou také nazývány frejmy) jsou typicky vytvořeny buď rozšířením existující báze, nebo sloučením různých bazí. Narozdíl od vektorů, které tvoří konečně rozměrné prostory, může být problém formulován i obecněji v rámci nekonečně rozměrných separabilních Hilbertových prostorů (Veselý, 2002b; Christensen, 2003). Tento funkcionální přístup nám umožňuje nacházet v těchto prostorech přesnější reprezentace objektů, které, na rozdíl od vektorů, nejsou diskrétní. V této disertační práci se zabývám hledáním řídkých representací v přeurčených modelech časových řad náhodných veličin s konečnými druhými momenty. Numerická studie zachycuje výhody a omezení tohoto přístupu aplikovaného na zobecněné lineární modely a na vícerozměrné ARMA modely. Analýzou mnoha numerických simulací i modelů reálných procesů můžeme říci, že tyto metody spolehlivě identifikují parametry blízké nule, a tak nám umožňují redukovat původně špatně podmíněný přeparametrizovaný model. Tímto významně redukují počet odhadovaných parametrů. V konečném důsledku se tak nemusíme starat o řády modelů, jejichž zjišťování je většinou předběžným krokem standardních technik. Pro kratší časové řady (100 a méně vzorků) řídké odhady dávají lepší predikce v porovnání s těmi, které jsou založené na standardních metodách (např. maximální věrohodnosti v MATLABu - MATLAB System Identification Toolbox (IDENT)). Pro delší časové řady (500 a více) obě techniky dávají v podstatě stejně přesné predikce. Na druhou stranu řešení těchto problémů je náročnější, a to i časově, nicméně výpočetní doba je stále přijatelná.

Signals recorded from the brain often show rhythmic patterns at different frequencies, which are tightly coupled to the external stimuli as well as the internal state of the subject. These signals also have transient structures related to spiking or sudden onset of a stimulus, which have a duration not exceeding tens of milliseconds. Further, brain signals are highly non-stationary because both behavioral state and external stimuli can change over a short time scale. It is therefore essential to study brain signals using techniques that can represent both rhythmic and transient components of the signal. In Chapter 2, we describe a multi-scale decomposition technique based on an over-complete dictionary called matching pursuit (MP), and show that it is able to capture both sharp stimulus-onset transient and sustained gamma rhythm in local field potential recorded from the primary visual cortex. Gamma rhythm (30 to 80 Hz), often associated with high-level cortical functions, has been proposed to provide a temporal reference frame (“clock”) for spiking activity, for which it should have least center frequency variation and consistent phase for extended durations. However, recent studies have proposed that gamma occurs in short bursts and it cannot act as a reference. In Chapter 3, we propose another gamma duration estimator based on matching pursuit (MP) algorithm, which is tested with synthetic brain signals and found to be estimating the gamma duration efficiently. Applying this algorithm to real data from awake monkeys, we show that the median gamma duration is more than 330 ms, which could be long enough to support some cortical computations.