|
|
Speaker Verification without Feature Extraction
Lukáč, Peter ; Rohdin, Johan Andréas (oponent) ; Mošner, Ladislav (vedoucí práce)
Speaker verification is a field that is still improving its state of the art (SotA) and tries to meet the demands of its use in speaker authentication systems, forensic applications, etc. The improvements are made by the advancements in deep learning, the creation of new training and testing datasets and various speaker recognition challenges and speech workshops. In this thesis, we will explore models for speaker verification without feature extraction. Inputting the models with raw speaker waveform simplifies the pipeline of the systems, thus saving computational and memory resources and reducing the number of hyperparameters needed for creating the features from waveforms that affect the results. Currently, the models without feature extraction do not achieve the performance of the models with feature extraction. By applying various techniques to the models we will try to improve the baseline performance of the current models without feature extraction. The experiments with SotA techniques improved the performance of a model without feature extraction considerably however we still did not achieve the performance of a SotA model with feature extraction. However, the improvement is considerable enough so that we can use the improved model in a fusion with feature extraction model. We also discussed the experimental results and proposed improvements that aim to solve discovered limitations.
|
|
|
Voice Conversion
Lukáč, Peter ; Glembek, Ondřej (oponent) ; Černocký, Jan (vedoucí práce)
The matter of this thesis is voice conversion. Voice conversion is taking speech of one speaker, that we call source speaker and transforming it into speech that sounds as the speech of another speaker, that we call target speaker. This is accomplished using voice conversion system described in this thesis. As the framework for speech analysis and synthesis, we are using tool called STRAIGHT that was predominantly used in Voice Conversion Challenge 2016. Our voice conversion system is based on spectral conversion using feed-forward neural network and parallel training.
|
|
|
Optimalizace vlastností kolagenních pěn z rybího kolagenu pro medicínské a veterinární použití.
Lukáč, Peter ; Grus, Tomáš (vedoucí práce) ; Rohn, Vilém (oponent) ; Matia, Ivan (oponent)
V průběhu projektu byly vyvinuty unikátní kolagenní pěny z kolagenu získaného z kůže sladkovodní ryby (kapr obecný, Cyprinus carpio). Pomocí síťování karbodiimidem byl překonán problém s nestabilitou kolagenní matrix z kolagenu získávaného z chladnokrevných živočichů při tělesné teplotě savců. Následně byly pěny impregnovány antibiotiky (gentamicin a vankomycin) a opětovně lyofilizovány, což je postup, který zajišťuje požadovanou koncentraci antibiotika bez rizika následného vymytí při dalších technologických krocích. Uvedený produkt je, na rozdíl od přípravků z nesíťovanéhokolagenu, stabilní i při sterilizaci gamma zářením. Finální sterilizovaný produkt byl testován in vivo na potkaním modelu infikované rány. Byla prokázána efektivita v léčbě potenciálně letální infekce Pseudomonas aeruginosa a kmene Stafylococcus aureus rezistentní k meticilinu (MRSA). Vzhledem k vysoké potřebě profylaxe a terapie infekcí pooperačních a jiných ran právě výše uvedenými polyrezistentními původci se jedná o slibný prostředek k budoucímu klinickému využití. Zkušenosti, které jsme získali v průběhu uvolnování ATB z kolagenních pěn budou v dalším vyvoji použity pro impregnaci zevní kolagenní vrstvy cévní protézy, čímž bychom mohli eliminovat jednu z největších nevýhod a rizik spojených s použitím umělých materiálu a tím je...
|
|
|
Speaker Verification without Feature Extraction
Lukáč, Peter ; Rohdin, Johan Andréas (oponent) ; Mošner, Ladislav (vedoucí práce)
Speaker verification is a field that is still improving its state of the art (SotA) and tries to meet the demands of its use in speaker authentication systems, forensic applications, etc. The improvements are made by the advancements in deep learning, the creation of new training and testing datasets and various speaker recognition challenges and speech workshops. In this thesis, we will explore models for speaker verification without feature extraction. Inputting the models with raw speaker waveform simplifies the pipeline of the systems, thus saving computational and memory resources and reducing the number of hyperparameters needed for creating the features from waveforms that affect the results. Currently, the models without feature extraction do not achieve the performance of the models with feature extraction. By applying various techniques to the models we will try to improve the baseline performance of the current models without feature extraction. The experiments with SotA techniques improved the performance of a model without feature extraction considerably however we still did not achieve the performance of a SotA model with feature extraction. However, the improvement is considerable enough so that we can use the improved model in a fusion with feature extraction model. We also discussed the experimental results and proposed improvements that aim to solve discovered limitations.
|
|
|
Voice Conversion
Lukáč, Peter ; Glembek, Ondřej (oponent) ; Černocký, Jan (vedoucí práce)
The matter of this thesis is voice conversion. Voice conversion is taking speech of one speaker, that we call source speaker and transforming it into speech that sounds as the speech of another speaker, that we call target speaker. This is accomplished using voice conversion system described in this thesis. As the framework for speech analysis and synthesis, we are using tool called STRAIGHT that was predominantly used in Voice Conversion Challenge 2016. Our voice conversion system is based on spectral conversion using feed-forward neural network and parallel training.
|
host ::
RSS.