|
|
Development of modern acoustic features quantifying hypokinetic dysarthria
Kowolowski, Alexander ; Zvončák, Vojtěch (referee) ; Galáž, Zoltán (advisor)
This work deals with designing and testing of new acoustic features for analysis of dysprosodic speech occurring in hypokinetic dysarthria patients. 41 new features for dysprosody quantification (describing melody, loudness, rhythm and pace) are presented and tested in this work. New features can be divided into 7 groups. Inside the groups, features vary by the used statistical values. First four groups are based on absolute differences and cumulative sums of fundamental frequency and short-time energy of the signal. Fifth group contains features based on multiples of this fundamental frequency and short-time energy combined into one global intonation feature. Sixth group contains global time features, which are made of divisions between conventional rhythm and pace features. Last group contains global features for quantification of whole dysprosody, made of divisions between global intonation and global time features. All features were tested on Czech Parkinsonian speech database PARCZ. First, kernel density estimation was made and plotted for all features. Then correlation analysis with medicinal metadata was made, first for all the features, then for global features only. Next classification and regression analysis were made, using classification and regression trees algorithm (CART). This analysis was first made for all the features separately, then for all the data at once and eventually a sequential floating feature selection was made, to find out the best fitting combination of features for the current matter. Even though none of the features emerged as a universal best, there were a few features, that were appearing as one of the best repeatedly and also there was a trend that there was a bigger drop between the best and the second best feature, marking it as a much better feature for the given matter, than the rest of the tested. Results are included in the conclusion together with the discussion.
|
|
|
Detection of selected audio events in a real environment
Kowolowski, Alexander ; Burget, Radim (referee) ; Přinosil, Jiří (advisor)
This work deals with methods for the detection of dangerous events, in this case gunshots, in a real environment. First of all, a testing and training database of sounds from the MIVIA database was created. In this database, the files were contained in six versions of signal-to-noise ratio, so the subsequent testing of the selected methods took place for the various shuffled files, and it was found that some methods are more accurate for cleaner recordings than others, but less accurate for more noisy ones. For the typical feature extraction from the input sound, the mel-frequency cepstral coefficients method was always used. In the thesis, the methods of support vector machines and ensemble of a number of weak classifiers are gradually tested on the created databases. These methods are then further optimized, for example by using statistical variables, and after optimization they achieve better results, as expected. In the work, two scripts were created, where one created a training database and on this data trained the classifier and the other created the test database, tested the selected classifier and obtained the results. The results are processed by confusion matrix and several proportional variables such as accuracy, sensitivity, specificity and others are calculated. These results are always listed in the relevant chapter of the thesis in the tables and column charts and are properly commented on.
|
|
|
Development of modern acoustic features quantifying hypokinetic dysarthria
Kowolowski, Alexander ; Zvončák, Vojtěch (referee) ; Galáž, Zoltán (advisor)
This work deals with designing and testing of new acoustic features for analysis of dysprosodic speech occurring in hypokinetic dysarthria patients. 41 new features for dysprosody quantification (describing melody, loudness, rhythm and pace) are presented and tested in this work. New features can be divided into 7 groups. Inside the groups, features vary by the used statistical values. First four groups are based on absolute differences and cumulative sums of fundamental frequency and short-time energy of the signal. Fifth group contains features based on multiples of this fundamental frequency and short-time energy combined into one global intonation feature. Sixth group contains global time features, which are made of divisions between conventional rhythm and pace features. Last group contains global features for quantification of whole dysprosody, made of divisions between global intonation and global time features. All features were tested on Czech Parkinsonian speech database PARCZ. First, kernel density estimation was made and plotted for all features. Then correlation analysis with medicinal metadata was made, first for all the features, then for global features only. Next classification and regression analysis were made, using classification and regression trees algorithm (CART). This analysis was first made for all the features separately, then for all the data at once and eventually a sequential floating feature selection was made, to find out the best fitting combination of features for the current matter. Even though none of the features emerged as a universal best, there were a few features, that were appearing as one of the best repeatedly and also there was a trend that there was a bigger drop between the best and the second best feature, marking it as a much better feature for the given matter, than the rest of the tested. Results are included in the conclusion together with the discussion.
|
|
|
Detection of selected audio events in a real environment
Kowolowski, Alexander ; Burget, Radim (referee) ; Přinosil, Jiří (advisor)
This work deals with methods for the detection of dangerous events, in this case gunshots, in a real environment. First of all, a testing and training database of sounds from the MIVIA database was created. In this database, the files were contained in six versions of signal-to-noise ratio, so the subsequent testing of the selected methods took place for the various shuffled files, and it was found that some methods are more accurate for cleaner recordings than others, but less accurate for more noisy ones. For the typical feature extraction from the input sound, the mel-frequency cepstral coefficients method was always used. In the thesis, the methods of support vector machines and ensemble of a number of weak classifiers are gradually tested on the created databases. These methods are then further optimized, for example by using statistical variables, and after optimization they achieve better results, as expected. In the work, two scripts were created, where one created a training database and on this data trained the classifier and the other created the test database, tested the selected classifier and obtained the results. The results are processed by confusion matrix and several proportional variables such as accuracy, sensitivity, specificity and others are calculated. These results are always listed in the relevant chapter of the thesis in the tables and column charts and are properly commented on.
|
guest ::
