|
|
Speech Technology Application in Pronunciation Training and Foreign Language Learning
Barotová, Štěpánka ; Žmolíková, Kateřina (referee) ; Szőke, Igor (advisor)
Tato diplomová práce pojednává o využití algoritmu Dynamic Time Warping (DTW) pro automatické hodnocení výslovnosti anglického jazyka. Práce se zaměřuje na vylepšení již existující aplikace pro výuku výslovnosti, a to ve třech oblastech: uživatelské rozhraní, samotný algoritmus a korektivní zpětná vazba uživateli. První část se věnuje přehledu technik používaných v této oblasti, následně je představen nový design uživatelského rozhraní, popsán navržený systém a experimenty. Experimenty se zaměřují na problematiku detekce chyb na úrovni fonémů, na detekci chyb v primárním důrazu na úrovni slabik a na hodnocení intonace na úrovni slov. Všechny použité metody jsou navrženy tak, aby poskytovaly korektivní zpětnou vazbu uživateli. V poslední části je popsáno, jak byly všechny tři vylepšené oblasti aplikace otestovány.
|
|
|
Adding New Words in a Dynamic Speech Recognition Decoder
Škrdlík, Kryštof ; Veselý, Karel (referee) ; Schwarz, Petr (advisor)
The result of this thesis is a modified speech recognizer developed by the company Phonexia, in which new words that are not part of its dictionary can be added dynamically. The chosen method that was implemented works by inserting finite automata with new words directly into a modified static recognition network describing combined language and pronunciation model of the recognizer in places that are specified in advance. This method offers comparable results with a speech recognizer without modification.
|
|
| |
|
|
Very limited Vocabulary Speech Recognizer
Vystavěl, Kamil ; Míča, Ivan (referee) ; Sysel, Petr (advisor)
This bachelor thesis deals with the implementation of voice diagnostic method with limited number of recognized words in Matlab environment. Recognizer is designed for recognition of isolated words and is based on the dynamic programming method. This method is realized by the dynamic time warping algorithm (DTW). Features of the speech signal are calculated by methods of short-term analysis in time and frequency domain and by methods that are based on cepstral analysis and linear predictive analysis. The representation of the word, which is generated from its features, is suitable for quantifying the degree of similarity with the representation of another word. In order to achieve the highest degree of similarity, the dynamic time warping algorithm eliminates influence of fluctuation of the speech rate by non-linear normalization time axis of one of the compared words. The degree of the similarity of the two compared words is enumerated as the words’ distance. The representations of known words are stored in a word-book. The unknown word is compared with all words in the word-book and recognizer calculates distances between every known word and the unknown word. The unknown word is defined as identical with the known word that has the shortest distance to the unknown word. The successfulness depends mainly on the choice of the features.
|
|
|
Voice Recording and Search for Skype
Nytra, Jiří ; Szőke, Igor (referee) ; Schwarz, Petr (advisor)
This work deals with the creation of a program communicating with Skype, which provides record calls in which can search for keywords by using advanced speech recognition technology. The work is presented and the interface protocol to communicate with Skype, call recording and method LVCSR for searching keywords.
|
|
|
Network Interface for Keyword Spotting System
Skotnica, Martin ; Glembek, Ondřej (referee) ; Szőke, Igor (advisor)
A considerable part of the research in computer science is dedicated to speech recognition as the speech-controlled systems become useful in many applications. One of them is the keyword spotting which makes possible to find words in audio data. Such a detector is developed at BUT Faculty of Information Technology. The goal of this work is to propose a network interface to this keyword detector based on client/server architecture. Client connects to the server and sends audio data. Server runs keyword detector with this received data and sends the result of keyword spotting back to client. Finally client visualizes the result and interact with user.
|
|
|
Visualization of User Pronunciations for Electronic Dictionarties
Pešán, Jan ; Chalupníček, Kamil (referee) ; Černocký, Jan (advisor)
The aim of this bachelor's work is to try to find a new way for development in learning capabilities of electronic dictionaries. There is an introduction of the main concept of learning pronunciations with visualization of phonemes in the first part. It is followed by chapter, which does a global review of methods for speech processing used in this project, e.g. HMM or Viterbi algorithm. In the third chapter, there is description of tools that we have used for implementation of the whole system. Next chapter explains more in detail technology of neural networks, used here as probability estimator. There is also a description of problem with compatibility of the used phoneme sets and in addition, it describes used phoneme models. Chapter 5 is whole about implementation of the system. There are also described scripts and tools applied for the preparation of the source data. In the next chapter, there is a user testing with screenshots. Moreover, in the last chapter I wrote a short conclusion and possible future ways for further developing of this system.
|
|
|
Penetration Tests of Speaker Verification System
Wojnar, Filip ; Landini, Federico Nicolás (referee) ; Plchot, Oldřich (advisor)
Cílem práce je provést penetrační testy na systému pro automatickou verifikace řečníka za použití syntézy řeči. Práce se zabývá fungování systému pro automatickou verifikaci řečníka a spoofing útoky na systémy, zabývající se touto problematikou. Práce se také podrobnějí zabývá fungováním syntézy řeči. Pozdější kapitoly se zabývají realizací penetračních testů a výsledky, které nám tyto testy přinesly.
|
|
|
Parallel Training of Neural Networks for Speech Recognition
Veselý, Karel ; Fousek, Petr (referee) ; Burget, Lukáš (advisor)
This thesis deals with different parallelizations of training procedure for artificial neural networks. The networks are trained as phoneme-state acoustic descriptors for speech recognition. Two effective parallelization strategies were implemented and compared. The first strategy is data parallelization, where the training is split into several POSIX threads. The second strategy is node parallelization, which uses CUDA framework for general purpose computing on modern graphic cards. The first strategy showed a 4x speed-up, while using the second strategy we observed nearly 10x speed-up. The Stochastic Gradient Descent algorithm with error backpropagation was used for the training. After a short introduction, the second chapter of this thesis shows the motivation and introduces the neural networks into the context of speech recognition. The third chapter is theoretical, the anatomy of a neural network and the used training method are discussed. The following chapters are focused on the design and implementation of the project, while the phases of the iterative development are described. The last extensive chapter describes the setup of the testing system and reports the experimental results. Finally, the obtained results are concluded and the possible extensions of the project are proposed.
|
|
|
Unsupervised Adaptation of Speech Recognizer
Švec, Ján ; Karafiát, Martin (referee) ; Schwarz, Petr (advisor)
The goal of this thesis is to design and test techniques for unsupervised adaptation of speech recognizers on some audio data without any textual transcripts. A training set is prepared at first, and a baseline speech recognition system is trained. This sistem is used to transcribe some unseen data. We will experiment with an adaptation data selection process based on some speech transcript quality measurement. The system is re-trained on this new set than, and the accuracy is evaluated. Then we experiment with the amount of adaptation data.
|
guest ::
