|
|
Estimation and analysis of speech signal periodicity
Malucha, Jan ; Poměnková, Jitka (referee) ; Sigmund, Milan (advisor)
Theoretical part starts with a short introduction to speech signals and short time analysis method. Concept of periodicity is clarified. This is followed by analysis of the speech signal parameters - voiced/unvoiced classification, intonation and short time period similarity. Next the overview of methods used to analyze the mentioned parameters together with concepts of their algorithms is provided. These methods include ZCR, STE, HNR, ACF, NCC, AMDF and DTW. Algorithms using the methods with supporting preprocessing algorithms were programmed in MATLAB for practical part. All were tested on real speech signals and the results are discussed at the end of the thesis.
|
|
|
Switching off of Electrical Tools by Voice
Rozsypálek, Lukáš ; Šebesta, Vladimír (referee) ; Sigmund, Milan (advisor)
In theoretical part Master’s thesis deals with processing acoustic signal before speech recognition. There are described methods of automatic speech recognition and capture attributes. Those methods are short-term energy of signal, short-term autocorrelation analysis, linear predictive analysis etc. In practical part has been created software, which has to switch off of electrical tool, if the keyword “zastav” has been spoken. In the second part of this thesis software has been optimized for work in real conditions (noise generated by electrical tools).
|
|
|
Speech denoising based on wavelet transform and voice recognition in segments
Chrápek, Tomáš ; Sysel, Petr (referee) ; Rajmic, Pavel (advisor)
The wavelet transform is a modern signal processing tool. The wavelet transform earned itself a great success mainly for its unique properties, such as the capability of recognizing very fast changes in processed signal. The theoretical part of this work is introduction to wavelet theory, more specifically wavelet types, a wavelet transform and its application in systems dealing with signal denoising. A main problem connected to speech signals denoising was introduced. The problem is degradation of the speech signal when denoising unvoiced parts. It is because of the fact that unvoiced parts and noise itself have very similar characteristics. The solution would be to apply different attitude to voiced and unvoiced segments of the speech. The main goal of this diploma thesis was to create an application implementing the speech signal denoising using the wavelet transform. The special attention should have been paid to applying different attitude to voiced and unvoiced segments of the speech. The demanded application is programmed as a grafical user interface (GUI) in MATLAB environment. The algorithm implemented in this form allows users to test introduced procedures with a great comfort. This work presents achieved results and discusses them considering general requirements posed on an application of given type. The most important conlusion of this Diploma Thesis is the fact that some kind of trade-off between sufficient signal denoising and keeping the speech understandable has to be made.
|
|
|
Pre-fortis shortening in Czech English
Hrychová, Helena ; Červinková Poesová, Kristýna (advisor) ; Pípalová, Renata (referee)
This bachelor thesis examines the role of a phenomenon called pre-fortis shortening in Czech English. This term is used by phoneticians to represent the shortening of vowels before fortis consonants. It is not characteristic only of English, Matthew Chen even suggested that it could be a language-universal phenomenon. In English it is a very significant feature since English vowel duration was proven by various experiments to serve as a cue to the perception of the voicing characteristic of the following consonant. This thesis advances our understanding of pre-fortis shortening in Czech English where vowel duration was hypothesized to correlate with the speaker's proficiency in English. As a result, this work aims to examine the extent to which 20 Czech students divided into two categories depending on their English level (A2, C1) exploit this phenomenon and whether their proficiency has an impact on vowel durations before fortis and lenis consonants. They were recorded while reading 40 English sentences and their vowel durations were compared to those obtained from two native English speakers.
|
|
|
Relationship between the strength of Czech accent and the duration of vowels before obstruents
Fejlová, Dita ; Skarnitzl, Radek (advisor) ; Tichý, Ondřej (referee)
This bachelor thesis aims to inspect the presence or absence of a feature called pre- fortis shortening in English spoken by Czechs. The term denotes the shortening of a vowel preceding a voiceless obstruent. This feature is known to appear in various languages like Russian, French, Italian; indeed, Matthew Chen even suggests it is language-universal. In English the feature is very prominent and because it affects speech perception, it is even considered a primary indicator of the voicing of the following obstruent. A study included in this thesis examines the extent to which 12 female speakers of Czech English, sorted into 3 categories according to their proficiency in pronunciation, mark the distinction between words like "bet" and "bed" by vowel shortening. The study does not exploit minimal pairs like these, but vowel-obstruent sequences taken from long read passages of BBC news. The the fortis/lenis character of the final obstruent contained in these sequences is a discerning parameter which always separates the data into two groups, the vowel durations of which are then compared. Statistical analysis of the data showed that contrary to the expectations, speakers with native-like pronunciation were not the ones who displayed the most massive usage of pre-fortis shortening. Possible reasons for...
|
|
|
Electropalatographic study of voicing assimilation in Czech alveolar plosives
Vaněčková, Kateřina ; Skarnitzl, Radek (advisor) ; Machač, Pavel (referee)
The phonological contrast of voicing is based on several phonetic differences which are yet to be satisfactorily explained. The most obvious of them is the presence of vocal cords vibration which has so far been considered crucial in Czech. For capturing the differences in other dimensions, the concept of the strength of articulation, or fortis/lenis contrast have been established. There are some rare cases in languages when the couple of phonemes with the voicing contrast differs also in another feature that cannot be explained by the fortis/lenis dimension correlates. This is the case of Czech alveolar plosives which have been described to differ in their place of articulation: the linguopalatal contact is more anterior for [t] that for [d]. The main aim of this study is to ascertain whether this contrast can indicate the original phoneme in the contexts where the voicing contrast is neutralized due to assimilation. Nevertheless, our electropalatographic data from 7 speakers of Czech show that not only the phonemes do not seem to influence the anteriority of the voicing realisation, but also the very asymmetry of linguopalatal contact in the voicing realisations tends to be neutralised in the segmental context used in our study, that is in coda position in the word-final syllable followed by the...
|
|
| |
|
|
Estimation and analysis of speech signal periodicity
Malucha, Jan ; Poměnková, Jitka (referee) ; Sigmund, Milan (advisor)
Theoretical part starts with a short introduction to speech signals and short time analysis method. Concept of periodicity is clarified. This is followed by analysis of the speech signal parameters - voiced/unvoiced classification, intonation and short time period similarity. Next the overview of methods used to analyze the mentioned parameters together with concepts of their algorithms is provided. These methods include ZCR, STE, HNR, ACF, NCC, AMDF and DTW. Algorithms using the methods with supporting preprocessing algorithms were programmed in MATLAB for practical part. All were tested on real speech signals and the results are discussed at the end of the thesis.
|
|
|
Pre-fortis shortening in Czech English
Hrychová, Helena ; Červinková Poesová, Kristýna (advisor) ; Pípalová, Renata (referee)
This bachelor thesis examines the role of a phenomenon called pre-fortis shortening in Czech English. This term is used by phoneticians to represent the shortening of vowels before fortis consonants. It is not characteristic only of English, Matthew Chen even suggested that it could be a language-universal phenomenon. In English it is a very significant feature since English vowel duration was proven by various experiments to serve as a cue to the perception of the voicing characteristic of the following consonant. This thesis advances our understanding of pre-fortis shortening in Czech English where vowel duration was hypothesized to correlate with the speaker's proficiency in English. As a result, this work aims to examine the extent to which 20 Czech students divided into two categories depending on their English level (A2, C1) exploit this phenomenon and whether their proficiency has an impact on vowel durations before fortis and lenis consonants. They were recorded while reading 40 English sentences and their vowel durations were compared to those obtained from two native English speakers.
|
|
| |
guest ::
