|
|
Quantitative Analysis of Vocal Tract Resonances in Patients with Parkinson’s Disease
Kováč, Daniel
Parkinson’s disease (PD) is often associated with hypokinetic dysarthria, impacting speech-motor function and resulting in articulatory deficits such as reduced speech intelligibility due to stiffness in articulatory structures. This study aims to introduce novel speech metrics to assess articulation impairments and differentiate between healthy individuals and those with PD. Resonant frequency attenuation (RFA) was confirmed as a valid measure, and two additional acoustic features based on supraglottic resonances were proposed. Their efficacy in distinguishing between groups was evaluated using a dataset comprising 19 healthy controls (HC) and 28 PD patients. Significantly divergent values between PD and HC were observed for all three features via Student’s t-test. Employing logistic regression with leave-one-out cross-validation, PD classification achieved 96% sensitivity and 100% specificity. This preliminary study suggests that supraglottic resonances could be pivotal in aiding PD diagnosis through acoustic speech analysis.
|
|
|
Analysis of temporal speech impairments in patients with Dementia with Lewy Bodies
Davaajargal, Anar ; Kováč, Daniel (referee) ; Novotný, Kryštof (advisor)
Dementia with Lewy bodies is the second most common neurodegenerative disease of the dementia type. Progressive degradation of motor and cognitive abilities and behavioral changes have a significant impact on the quality of life of the affected person and those around them. Early diagnosis of the disease is therefore crucial for setting up adequate treatment. Due to similarities with other diseases, it may be overlooked or mistaken for another disease in the early stages. There is a lack of extensive research detailing the speech defects specific for this disease, which through acoustic analysis could be used to make a paraclinical diagnosis in a cost-effective, non-invasive, yet efficient manner. The aim of this paper is to survey available knowledge and methods, select and implement appropriate parameters targeting tempo and pauses, and then evaluate the effectiveness of the methods used to discriminate patients from healthy controls. The evaluation was based on Mann-Whitney U test and descriptive statistics. The best results were obtained for parameters MWPM of longer words, median pause length, occurrence of repeated words and NSR, all considering the task with monologue, which generally showed a greater discriminative power than the task that involved reading a text. The analysis suggests temporal parameters suitable for describing speech disorders in the selected disease. At the same time, new custom metrics are introduced and tested in this paper.
|
|
|
Acoustic analysis of emotionally affected sentences in patients with Parkinson's disease
Gavlasová, Radka ; Kováč, Daniel (referee) ; Mekyska, Jiří (advisor)
This thesis focuses on Parkinson's disease and its effect on emotional expression in speech. The aim was to conduct a literature search on acoustic emotional analysis of PD patients and to implement acoustic parameters to distinguish between healthy and diseased individuals. The database used contained recordings of 100 patients with PD and 52 healthy controls for various speech tasks. For this analysis, 7 emotionally coloured sentences and 11 acoustic parameters were selected and implemented in Python. From the statistical analysis, it was found that the most significant parameters include pauses in speech and intensity variability. The XGBoost algorithm with 10-fold stratified cross-validation was used for classification. A total of 10 models were implemented to analyze all tasks together and each task separately. Optimization was performed using randomized search. For the combination of all tasks, the significant parameter was the variability in intensity or speech rate. For the individual speech tasks, variability in intonation and formant areas was highly significant. The best model achieved a 63% success rate (BACC) and 85% sensitivity. The results suggest that emotional prosody affects classification, confirming previous findings and pointing to the need for further investigation in this area.
|
|
|
Automated segmentation of the diadochokinetic task for remote monitoring of hypokinetic dysarthria
Svojanovský, Jan ; Mekyska, Jiří (referee) ; Kováč, Daniel (advisor)
The study describes health problems associated with Parkinson’s disease, especially hypokinetic dysarthria. It also points out the subjective and objective methods used to determine the severity of the disease. One of these methods is a diadochokinetic (DDK) task based on rapid syllable repetition to test the functionality of the articulatory apparatus (e.g., tongue, lips, or vocal cords). Correct speech production can also be examined by a speech therapist in the 3F test, which scores the severity of disorders in different areas of speech production. Next, the approaches of other authors, also dealing with the automated search of syllables in the speech signal, are described. The thesis also discusses some features of human speech that are needed for training a machine learning model. These features were computed for each of the 30 ms segments of a DDK task. The main goal is the automated detection and classification of [Pa]-[Ta]-[Ka] syllables in the recordings. For this purpose, an algorithm using a logistic regression was applied. The resulting average accuracy of syllable detection in the recordings was 89.4 %, average sensitivity 59.0 % and average specificity 93.79 %. The identification of individual syllable types was successful with an average accuracy of 90.78 %, an average sensitivity of 59.0 % and an average specificity of 95.39 %. Considering that the predicted onset was not located directly on the manually annotated onset, but in its close vicinity (up to ±3 segments), the average detection sensitivity and average syllable type classification sensitivity were 96.9 % and 85.1 % respectively, with an average difference between manually annotated and automatically segmented syllable onsets of 10.35 ms. The average accuracy of classification of speakers into healthy and PN patients using logistic regression (with speech parameters obtained after automated segmentation) was only 43.92 %, sensitivity 70.0 % and specificity 30.61 % (threshold 70 %). Using linear regression, the clinical scores of the 3F test were predicted. For faciokinesis, the root mean square error (RMSE) was 2.764 after manual syllable annotation and 3.271 after automated segmentation. The RMSE values for phonetics were 3.657 (manual) and 0.753 (automated). The developed algorithm can detect syllables in DDK tasks with relative success, and thus it is possible to determine parameters quantifying speech disorders with low differences with manual segmentation. If the recordings of DDK tasks meet the conditions for computing all these parameters, the algorithm could be used to classify speakers into healthy subjects and PN patients, for whom it could additionally assess the severity of dysarthria.
|
|
|
Analysis of impact of noise in recordings on the automated detection of hypokinetic dysarthria
Havelková, Nikola ; Galáž, Zoltán (referee) ; Kováč, Daniel (advisor)
This thesis deals with the automated detection of hypokinetic dysarthria by analysing the influence of noise present in recordings. Appropriate single-channel methods, specifically the spectral subtraction and Kalman filter, are selected and implemented in the MATLAB R2022a to enhance speech. These methods are also used for noise-free recordings, to which additive white noise was added. Afterwards, the effectiveness of these methods is objectively evaluated by using signal-to-noise ratio values. After enhancing of speech, interferences are extracted from the recordings. The effect of the presence of noise, as well as its subsequent suppression by individual methods, is then evaluated by statistical analysis, specifically using the Kruskal-Wallis test and the post hoc Dunn’s test. The probability of distributing parameters of clean, noisy and enhanced recordings, for which the effect of noise is significant, according to statistical tests, are plotted using violin and box graphs. Finally, the classification was done by logistic regression with the help of machine learning, where the effect of the presence of noise and subsequent speech enhancement on automated detection of hypokinetic dysarthria was described according to the area values under the ROC curve.
|
|
|
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 %.
|
|
|
Automatic speech recordings segmentation tool
Santa, Roman ; Zvončák, Vojtěch (referee) ; Kováč, Daniel (advisor)
Nástroj pre automatickú segmentáciu spracováva nahrávky reči a extrahuje hovorené slovo z nahrávok. Je dôležité, aby pokročilá analýza pracovala iba s rečovými časťami z nahrávky. Nástroj na segmentáciu má ulahčiť spracovanie nahrávok pre analýzu rozdielov medzi hláskami pacientov s parkinsonovou chorobou a tými zdravými. Cieľ tejto práce je navrhnúť a otestovať detektory reči s Google WebRTC detektorom a vybrať ten najvhodnejší detektor reči s minimálnym počtom chýb. Ďalej, vytvoriť nástroj na segmentáciu nahrávok a otestovať rozpoznávanie reči pomocou dynamic time warping. Bola použitá databáza poskytnutá laboratóriom pre analýzu mozgových ochorení. Obsahuje české a maďarské nahrávky s rovnakým počtom mužských a ženských pacientov a aj rovnakým počtom zdravých pacientov a pacientov s parkinsonovou chorobou. Najlepšie výsledky v testoch dosiahol detektor na základe energie reči. Nebol zistený žiaden rozdiel v presnosti detektoru pri spracovaní mužských a ženských nahrávok alebo nahrávok zdravých či chorých pacientov. Nahrávky s nízkym odstupom signálu od šumu boli náročnejšie na spracovanie s frekvenciou chýb od 12%. Na základe výsledkov, bol navrhnutý nový detektor pre spracovanie úplnej nahrávky. Na záver bol testovaný algoritmus pre rozpoznávanie podobnosti reči na základe melovských kepstrálnych koeficientov.
|
|
|
Bridge over drain channel
Kováč, Daniel ; Nečas, Radim (referee) ; Panáček, Josef (advisor)
This bachelor thesis is foucused on design monolithic road bridge. The bridge is situated on Road R1 between Žarnovica and Šášovské Pdhradie. The primary funkcion of bridge is crossing open drain channel, secondary it is biocorridor. There are three preliminary versions designed, two of them are using monolitic prestressed concreate beams. One version is using precast beams. The project includes statical analysis drawing documentation.
|
|
|
3D Audio Standards in Home Theatre Environment
Kováč, Daniel ; Schimmel, Jiří (referee) ; Balík, Miroslav (advisor)
This work deals with object-based sound. There is a description of the difference between channel-based and object-based sound and the development of this sound to today's form. Subsequently, formats that work with objects are described, and they are compared with each other. MPEGH-3D Audio format is specified here, which is a standard today, and also the method of positioning the virtual audio source by vector-base amplitude panning. It also describes how an encoder and decoder supporting speaker configurations 5.1.2, 5.1.4 and 7.1.4 has been developed according to the MPEG-H 3D Audio specification. Lastly, the work deals with the problem of panning the object sound to the standard configuration and there is also a mention of the listening test.
|
|
|
Bridge over a local road
Kováč, Daniel ; Mojzík, Petr (referee) ; Nečas, Radim (advisor)
This master thesis is focused for design, respectively for creating counterproposal of existing structure of Motorway Bridge located at D1 motorway between Jánovce and Jablonov in district Levoča in Slovakia. This bridge is over a local road. The main aim of this thesis is design, analysis and assessment of bearing elements of bridge deck. For purposes of this thesis was created three variants of bridge deck. Detailed analysis was performed on bearing construction from post-tensioned precast beam with monolithing concrete slab. Two mathematic models were used for analysis of chosen variant. First model was spatial 3D shell construction. The purpose of this model was determinate cross spreading line of other constant load and live load. Second, flat 2D model, was for determination creep and shrinkage appeared from long-term load. These effects were observed in at advanced defined time nodes. Outcomes from upper mentioned model were used from design and assessment of bearing elements of bridge deck, design prestressed reinforcement cables and concrete reinforcement rods.
|
guest ::
