|
| |
|
|
Correlation analysis of Parkinson's disease in the acoustic field
Vošček, Jakub ; Mekyska, Jiří (referee) ; Smékal, Zdeněk (advisor)
The topic of this bachelor thesis is the correlation analysis of Parkinson’s disease in the acoustic field. The first part is about the Parkinson disease and its symptoms. It looks closer on problems with speech production, which is called hypokinetic disarthria, and describes the causes of the problems as well as the kind of treatment that is used. The next part involves a study of the pre–processing of signal, i.e. removing the direct component, a preemphasis and a segmentation to smaller frames. Afterwards, individual parameters are calculated in the next step. It is also necessary to calculate simple statistics, for example a median, a standard deviation, etc. after the calculation of some parameters. The calculation of Pearson’s and Spearman’s correlation coefficients is included. Moreover, a block diagram for the data processing is suggested, which involves a description of the functions of the individual blocks. The program is explained in the practical part, which also features parts of the tables with parameters' values and the calculated coefficients. As the conclusion of the work, there are graphs which display correlations of the parameters and the paraclinical data.
|
|
|
Identification of persons via voice imprint
Mekyska, Jiří ; Atassi, Hicham (referee) ; Smékal, Zdeněk (advisor)
This work deals with the text-dependent speaker recognition in systems, where just a few training samples exist. For the purpose of this recognition, the voice imprint based on different features (e.g. MFCC, PLP, ACW etc.) is proposed. At the beginning, there is described the way, how the speech signal is produced. Some speech characteristics important for speaker recognition are also mentioned. The next part of work deals with the speech signal analysis. There is mentioned the preprocessing and also the feature extraction methods. The following part describes the process of speaker recognition and mentions the evaluation of the used methods: speaker identification and verification. Last theoretically based part of work deals with the classifiers which are suitable for the text-dependent recognition. The classifiers based on fractional distances, dynamic time warping, dispersion matching and vector quantization are mentioned. This work continues by design and realization of system, which evaluates all described classifiers for voice imprint based on different features.
|
|
| |
|
|
Identification of sleep disorders based on actigraphy data and sleep diaries
Molík, Miroslav ; Mekyska, Jiří (referee) ; Mikulec, Marek (advisor)
This master’s thesis deals with prediction of Parkinson's disease using sleep parameters from actigraphy and sleep diaries. The goal is to design a machine learning approach, which will be able to recognize pacients suffering from Parkinson's disease. For training dataset supplied by St. Anne's University Hospital Brno was used, which was variously modified for achieving best possible results. These adjustments were evaluated according to the results of the trained models and based on these results, two models (achieving test accuracies of 85 and 82%) were selected.
|
|
|
Speech Enhancement using Cancelling of Dissonant Components
Studený, Radim ; Mekyska, Jiří (referee) ; Smékal, Zdeněk (advisor)
This work is supposed to remove interferences from speech signal and so increase its clarity, quality of degraded signal and signal-to-noise ratio. The most common sources of interference could be street noise, wind coming on a microfon, speech on the background or music.The metod described in this work remove frequence bands of a signal, which in relation to the fundamental frequency of a speech are disonant. Be specific, to reference tone C there are F#, B a C# tones.
|
|
|
Functional connectivity and brain structure assessment in patients at risk of synucleinopathies
Klobušiaková, Patrícia ; Gajdoš, Martin (referee) ; Mekyska, Jiří (advisor)
Synucleinopathy is a neurodegenerative disorder characterized by the presence of pathological protein -synuclein in neurons. So far, treatment that could heal or permanently stop this disease is not known. The aim of this work is to identify prodromal stages of synucleinopathies using functional connectivity processed applying graph metrics and assessing cortical thickness and subcortical structures volumes from magnetic resonance imaging data, and to verify specificity and sensitivity of combinations of parameters that sufficiently differentiate patients in risk of synucleinopathies. To accomplish this goal, we collected data from patients in the risk of synucleinopathy (preDLB, n = 27) and healthy controls (HC, n = 28). We found reduced volume of right pallidum and increased hippocampal volume to cortical volume ratio, increased normalised clustering coefficient and higher modularity in the preDLB group in comparison to HC. These four parameters were modeled using machine learning. The resulting model differentiated preDLB and HC with balanced accuracy of 88 %, specificity of 89 % and sensitivity of 86 %. The findings of this thesis can serve as the basis for further studies searching for specific MRI markers of prodromal stage of synucleinopathy that could be targeted with therapy in the future.
|
|
|
Automated diagnosis of sleep disorders using wearable devices
Sigmund, Jan ; Mekyska, Jiří (referee) ; Mikulec, Marek (advisor)
Sleep disorders induce many negative repercussions. Furthermore, research about their connection to cognitive health is increasing in numbers. This thesis concerns detection of poor sleep quality via raw actigraphy data. Existing method for assessing sleep was selected, it’s performance was validated against polysomnography on 27 patients. Used algorithm defines sleep as the absence of change in arm angle. Resulting 81 % sensitivity, 62 % specificity and 78 % accuracy is different from the outcome in the pilot study. Two approaches, to determine sleep quality were used. Both are based on comparing sleep features – first, with National Sleep Foundation recommendations and second, with control group without sleep disorders (7 persons). The goal was to pinpoint the remaining 19 patients with diagnosis. The recommendation for SOL, WASO, NA>5 and SE had higher sensitivity (75 %), lower specificity (71 %) and identical accuracy (74 %). These approaches were then also tested on 7-day actigraphy, consisting of 27 subjects, that are presumed to have prodromal dementia with Lewy bodies. Same principle was applied to try to predict LBD and thereby address the link between sleep quality and neurodegeneration. This resulted in 86 % sensitivity, 38 % specificity and 63 % accuracy. With regard to achieving solid sensitivity in all cases and good accuracy this could be used to indicate sleep quality.
|
|
|
Differential analysis of multilingual corpus in patients with neurodegenerative diseases
Kováč, Daniel ; Zvončák, Vojtěch (referee) ; Mekyska, Jiří (advisor)
This diploma thesis focuses on the automated diagnosis of hypokinetic dysarthria in the multilingual speech corpus, which is a motor speech disorder that occurs in patients with neurodegenerative diseases such as Parkinson’s disease. The automatic speech recognition approach to diagnosis is based on the acoustic analysis of speech and subsequent use of mathematical models. The popularity of this method is on the rise due to its objectivity and the possibility of working simultaneously on different languages. The aim of this work is to find out which acoustic parameters have high discriminative power and are universal for multiple languages. To achieve this, a statistical analysis of parameterized speech tasks and subsequent modelling by machine learning methods was used. The analyses were performed for Czech, American English, Hungarian and all languages together. It was found that only some parameters enable the diagnosis of the hypokinetic disorder and are, at the same time, universal for multiple languages. The relF2SD parameter shows the best results, followed by the NST parameter. When classifying speakers of all the languages together, the model achieves accuracy of 59 % and sensitivity of 72 %.
|
|
|
Speech signal processing via web interface
Markovič, Michal ; Smékal, Zdeněk (referee) ; Mekyska, Jiří (advisor)
This work is resolving the possibility of processing a speech signal by using the web interface. A record of the speech signal is made and then saved and stored in servers. Then it’s possible to analyse and process the record with the help of a web application written in the C++ programming language. The results are shown in website.
|
guest ::
RSS feed.