|
|
Establishing speaker's age and sex
Rendek, Tomáš ; Pfeifer, Václav (referee) ; Atassi, Hicham (advisor)
This work deals with speaker´s age and gender recognition. At the beginning it introduces the practical usage of this application and discusses the solutions available. The theoretical part of the thesis specifies the feature extraction and reduction methods and speech databases used in the experiments. The practical part describes the recognizer implemented in the Emotional tool and in two chapters describes the individual experiments. Regarding speaker´s gender estimation; we focused on the impact of the emotional state and speaker's age on the classification process. The two remain experiments were dedicated for general gender estimation performed by using two different classifiers – GMM and k-NN. These two classifiers were used in age estimation as well. In this case, four Group of age was formed and two different feature sets namely: segmental and suprasegmental were exploited four groups
|
|
|
Multilingual analysis of human emotional states
Rendek, Tomáš ; Koula, Ivan (referee) ; Atassi, Hicham (advisor)
This work deals with the properties of the speech signal. At the beginning it introduces a process of generation of the speech. Then, it covers the prosodic features of the speech, which represent a related characteristic of emotions. It defines an emotion itself, as well as the basic features and parameters of the human speech. For the analysis we use the program called Praat. As it is an unknown program, we devote a part of the work to it, which acquaints us with its advantages. The next part of this paper comprises also two enclosed databases containing records of particular emotional states of human. These databases were created and collected for Slovak and German language. However, none of them contain spontaneous material. Next, the work concerns a concept of the neural networks. It regards it as a possible realization of recognizing of emotional characteristics. The initial analysis presents large number of gained features, out of which only the best twelve were selected on the basis of geometric separability. These features are distinct for both sexes, as well as for both nationalities. Consequently, they are used for training with a neural network. The work concludes by summarizing of the results discussing the successfulness with recognition of emotional states. It also gives possible reasons which lead to degradation of their successful classifying. The thesis contains a CD with all the partial and ultimate results, and files with records for Slovak and German language.
|
|
|
Phonetic processing of speech using Praat
Kráčala, Martin ; Staněk, Miroslav (referee) ; Sigmund, Milan (advisor)
The goal of this bachelor thesis is to create a Czech manual for beginners with software Praat. Praat is a software package designed for speech processing. It can be used for speech analysis and synthesis. In this work is described Praat user interface, operations with files in the program, sound recording, editing and analysis of these sounds. Manual is supplemented with examples of solved phonetic problems. The second goal of this thesis is to create a script to extract vowels and some consonants from voice records. The thesis contains description of chosen principles of recognition of vowels based on formants and intensity contour and the principle of recognition of consonants based on the ratios of energy in predetermined frequency bands. The success rate of the script is thoroughly examined and various optimization methods are discussed.
|
|
|
Application for the calculation of speech features describing hypokinetic dysarthria
Hynšt, Miroslav ; Mekyska, Jiří (referee) ; Kiska, Tomáš (advisor)
This thesis is about design and implementation of application for computing speech parameters on people with Parkinson disease. At the beginning is generaly described Parkinson disease and Hypokinetic dysarthria and how it affects the speech and speech parameters when it occurs. Mainly there are described areas of speech like phonation, prosody, articulation and fluent speech. As a part of next topic this thesis describes specific speech parameters with bigger meaning during diagnosis Parkinson disease and it's progress over the time. There are also mentioned few significant studies dealing with examination of speech of the subjects with diagnoses of Parkinson disease and computing some speech parameters in order to analyze their speech impairments. Part of the thesis is description of implemented standalone application for calculating, exporting and visualizing of speech parameters from selected sound records.
|
|
|
Comparison of analysis of speech in dependence on age and gender
Báňa, Josef ; Smékal, Zdeněk (referee) ; Atassi, Hicham (advisor)
This thesis deals with analysis of speech signal in dependence on the gender and the age of the speaker. We tried to investigate through the features to find the best set for the automatic classification of speakers. It also contains a brief discussion about the speech signal and its characteristics. We used a program called Praat for the speech analysis purpose. This program is also described in this work. We mainly focused on the suprasegmental features of speech. Our first step was to make our own speech corpus which should contain speech records from speakers with various age and gender. We made the analysis using Praat and reported it within this thesis. For the automatic classification purpose, twelve features were selected basing on there quality criteria and used with a neural network to classify the speakers to classes with different age and gender. As it was mentioned, a neural network was used as a classifier. We used “Neural Network Toolbox” in the Matlab program to create and train our networks.
|
|
|
Estimation of formant frequencies using machine learning
Káčerová, Erika ; Galáž, Zoltán (referee) ; Mekyska, Jiří (advisor)
This Master's thesis deals with the issue of formant extraction. A system of scripts in Matlab interface is created to generate values of the first three formant frequencies from speech recordings with the use of Praat and Snack(WaveSurfer). Mel Frequency Cepstral Coefficients and Linear Predictive Coefficients are extracted from the audio files in order to be added to the database. This database is then used to train a neural network. Finally, the designed neural network is tested.
|
|
|
Estimation of formant frequencies using machine learning
Káčerová, Erika ; Galáž, Zoltán (referee) ; Mekyska, Jiří (advisor)
This Master's thesis deals with the issue of formant extraction. A system of scripts in Matlab interface is created to generate values of the first three formant frequencies from speech recordings with the use of Praat and Snack(WaveSurfer). Mel Frequency Cepstral Coefficients and Linear Predictive Coefficients are extracted from the audio files in order to be added to the database. This database is then used to train a neural network. Finally, the designed neural network is tested.
|
|
|
Application for the calculation of speech features describing hypokinetic dysarthria
Hynšt, Miroslav ; Mekyska, Jiří (referee) ; Kiska, Tomáš (advisor)
This thesis is about design and implementation of application for computing speech parameters on people with Parkinson disease. At the beginning is generaly described Parkinson disease and Hypokinetic dysarthria and how it affects the speech and speech parameters when it occurs. Mainly there are described areas of speech like phonation, prosody, articulation and fluent speech. As a part of next topic this thesis describes specific speech parameters with bigger meaning during diagnosis Parkinson disease and it's progress over the time. There are also mentioned few significant studies dealing with examination of speech of the subjects with diagnoses of Parkinson disease and computing some speech parameters in order to analyze their speech impairments. Part of the thesis is description of implemented standalone application for calculating, exporting and visualizing of speech parameters from selected sound records.
|
|
|
Phonetic processing of speech using Praat
Kráčala, Martin ; Staněk, Miroslav (referee) ; Sigmund, Milan (advisor)
The goal of this bachelor thesis is to create a Czech manual for beginners with software Praat. Praat is a software package designed for speech processing. It can be used for speech analysis and synthesis. In this work is described Praat user interface, operations with files in the program, sound recording, editing and analysis of these sounds. Manual is supplemented with examples of solved phonetic problems. The second goal of this thesis is to create a script to extract vowels and some consonants from voice records. The thesis contains description of chosen principles of recognition of vowels based on formants and intensity contour and the principle of recognition of consonants based on the ratios of energy in predetermined frequency bands. The success rate of the script is thoroughly examined and various optimization methods are discussed.
|
|
|
Multilingual analysis of human emotional states
Rendek, Tomáš ; Koula, Ivan (referee) ; Atassi, Hicham (advisor)
This work deals with the properties of the speech signal. At the beginning it introduces a process of generation of the speech. Then, it covers the prosodic features of the speech, which represent a related characteristic of emotions. It defines an emotion itself, as well as the basic features and parameters of the human speech. For the analysis we use the program called Praat. As it is an unknown program, we devote a part of the work to it, which acquaints us with its advantages. The next part of this paper comprises also two enclosed databases containing records of particular emotional states of human. These databases were created and collected for Slovak and German language. However, none of them contain spontaneous material. Next, the work concerns a concept of the neural networks. It regards it as a possible realization of recognizing of emotional characteristics. The initial analysis presents large number of gained features, out of which only the best twelve were selected on the basis of geometric separability. These features are distinct for both sexes, as well as for both nationalities. Consequently, they are used for training with a neural network. The work concludes by summarizing of the results discussing the successfulness with recognition of emotional states. It also gives possible reasons which lead to degradation of their successful classifying. The thesis contains a CD with all the partial and ultimate results, and files with records for Slovak and German language.
|
guest ::
