What is the 'Saarbruecken Voice Database'? A collection of voice recordings from more than 2000 persons. One recording session contains the following recordings:
For the components listed above, the speech signal and the EGG signal have been stored in separate files. Any comments about the recordings are contained in an associated text-file. Depending on the quality of the recording, there are some recording sessions where not every vowel in every version is present.
This webinterface gives you acces to the 'Saarbruecken Voice Database'. It contains several internet pages, which can be used to
Before the individual internet pages are explained, some general tips are given here:
There is a special button on the startpage. You can use it to switch between the English and the German versions of the websites.
To have access to the 'Saarbruecken Voice Database', cookies have to be activated in your browser for safety reasons (see settings of your browser). Only one cookie with a random value will be saved on your computer. This value enables the server to generate dynamically according to your input (e.g. search criteria).
Furthermore, a username is needed. You can get this from the administrator (see the e-mail address at the foot of the page). After entering your username and password, click on the button 'Accept'. Your data should be transfered to the server via HTTPS (i.e. encoded). If the server does not accept the data, an error message appears. If the login was successful, the main menu appears.
With a successful login, amongst other things your access data will be saved temporarily on the server. This makes new input redundant. For safety reasons the button 'Logout' appears on each of the subsequent webpages. If you click on that button, all the information which has been saved temporarily on the server (since your last login), for example your password, will be deleted. If you have not clicked on any links or buttons for a while (depending on the server), the server also automatically deletes all your temporary records, for safety reasons.
Search criteria (e.g. a particular pathology) can be selected on this page. The criteria must be confirmed by clicking on the 'Accept' buttons with your mouse. The recording sessions that have been found will be listed below the table with the search criteria. The search criteria can be changed at any time. To update the list of recording sessions found, click on 'Accept' again.
The buttons 'Reset' reset all search criteria to a predefined state. These criteria are met by all recording sessions. This default applies automatically when you first open this page. The chapter Export contains important tips about the 'Reset' buttons.
'Export' buttons take you to a page where you can choose which of the recording sessions that have been found and marked should be downloaded (see chapter Results and chapter Export). Apart from this, these buttons also have the same functions as the 'Accept' buttons. That means after entering the search criteria (see following chapter) you can confirm the input directly with the 'Export' button.
The search criteria can be found in a number of grey-shaded table cells. You can search for recording sessions by using the criteria 'speaker' and 'recording session'. As long as there is at least one non-empty line with 'pathologies and diagnosis', the criteria defined in that line have to be met. In the following, the table cells with these three types (speaker, recording session, pathology and diagnosis) will be called "search criteria lines". Except for the sex of the speakers and the type of the recordings session, all criteria in one search criteria line must be met at the same time. It will be explained below when the criteria of a line are met by a recording session. Search criteria lines that are left blank do not affect the number of recording sessions found.
If one gender is selected, only recording sessions of speakers with this gender will be sought. If no gender is selected, this criterion will be irrelevant for the search in the database. The search can be limited to speakers of certain age groups. The following table exemplifies how an age group can be defined. Different age groups are separated by a comma (see figure below). Blanks have no effect. If you contravene the syntax rules, a warning note will appear when you click on 'Accept'. The age groups after the first syntax error will be deleted from the input box.
||Speakers who are 21 years old.
||Speakers up to 15 years old.
||Speakers who are at least 15 years old.
||Speakers who are at least 54 years old and 68 at the most.
To preserve the anonynimty of the speakers, each speaker has got a unique number. With the last box of the search criteria lines, recording sessions by a particular speaker can be defined. The syntax is the same as for the age of the speakers.
Search for recording sessions by female speakers who are between 40 and 50 years old:
Search for the first 100 recording sessions (relating to the number of all registered speakers in the database) by speakers who are female and between 40 and 50 years old:
For the type of the recording session things are organised similarly to the sex of the speaker.
Each recording session has got a clear number. Like the identification number of the speaker, the search can be limited to certain recording sessions. The syntax is the same as for the age of the speaker.
See examples under speaker.
Pathologies and Diagnosis
In this search criteria line, pathologies which a recording session must have (obligatory pathologies) or must not have (excluded pathologies) are defined. To do this, mark pathologies with your mouse and move them between the lists with the blue arrow buttons. All the available pathologies are listed under 'selection'.
Search for recording sessions with the pathology 'Vox senilis':
Search for recording sessions with the pathologies 'Laryngitis' and 'Vox senilis':
Search for recording sessions with 'Laryngitis' and 'Vox senilis' that occur without the accompanying pathology 'Dysphonie'.
With the button 'Anfrageerweiterung' you can expand the search criteria for an additional search criteria line of the type 'Pathologies and Diagnosis'. This enables you to seek recording sessions which have e.g. the pathology 'Laryngitis' or the pathology 'Vox senilis'. Each of the pathologies will be moved into a different search criteria line in the list 'obligatory pathologies'. A database request also delivers the recording sessions which met the criteria of one line as well as the other. The criteria of the two lines are not mutually exclusive.
A search criteria line that is no longer needed can be deleted with the button 'Delete'. An empty search criteria line will be ignored at the time of the database request.
In the input box 'Remarks w.r.t. diagnosis', any keyword can be entered. Upper and lower case are irrelevant. Recording sessions will be sought that meet all optional pathology criteria at the same time and that provide the entered keyword (see example 5 and 6). There are no conventions for the remarks text. A word like "chronic" can appear in abbreviated form. The remarks are in german.
Search for recording sessions with chronic Laryngitis:
Search for recording sessions with chronic Laryngitis or distinct hyperfunctional dysphonia:
Sorting is carried out according to the conditions specified. The preset conditions group the recording sessions by speaker. The recording sessions by one speaker are sorted chronologically.
After entering the search criteria and clicking on one of the 'Accept' buttons, a table with recording sessions that meet all of the selected criteria appears below the search criteria (see picture).
In the first column of the table, a mark for each recording session can be set or deleted. A set mark means, the file of this recording session will be downloaded when one of the 'Export' buttons is clicked. In a database request, all recording sessions that have been found are initially marked by default for export. If the search criteria are later changed, the marks
Irrespective of the recording for which the status of the marks has been saved, only those recording sessions are exported that meet the current search criteria and are marked. The 'Reset' buttons delete (among other things) the status of all marks. At the next database request, all recording sessions will be marked again.
In the left corners of the table there are the links 'Select' and 'Delete selection'. The former marks all listed recording sessions and the latter deletes all these marks. A mark can be set or deleted by clicking the squares in the first column of the table.
The second column contains the identification number of the recording session. Each number is linked to a page which shows information about all recordings by the speaker of this recording session. Via this page, the speech signal files from the recording sessions can be played directly (see section Speaker). More precisely, each of the identification numbers is linked to the corresponding recording session within the mentioned internet page.
The next two columns contain the type of the recording sessions and the date of the recording. Next, there is the identification number of the speaker. This is a link to the beginning of the internet page mentioned above, which is explained in the section Speaker. The subsequent table columns contain:
The specified pathologies of all search criteria lines are highlighted red in this table, together with the keywords that are sought in the column 'Remarks w.r.t. diagnosis'.
A maximum of ten recording sessions are listed per page. In the upper right hand corner, you can see how many recording sessions overall meet the criteria and which recording sessions are listed on the current page. The recording sessions can be browsed through with the arrow buttons. You can speed up the browsing by selecting a certain page. The page number must be confirmed by clicking one of the 'Accept' buttons.
On this page, all recording sessions by a certain speaker are listed chronologically. The speech signals of these recording sessions can be downloaded individually in WAV format or played directly.
At the head of the page there are two buttons. You can browse to the previous or the following speaker with these arrow buttons. The browsing only applies to speakers for which at least one recording session meets the current search criteria. The order of the speakers depends, on the one hand, on the sorting order of the database and, on the other hand, on the position of the first recording session found for each speaker.
A table is shown for each of the recording sessions (see picture). The blue or the grey background colour in the table show whether or not the recording session meets the search criteria.
The symbol to the left of the first line in this table shows the mark of the corresponding recording session. Next to that, there is the identification number of the recording session. The buttons with the loudspeaker symbol are directly linked to the speech signal files of the recording session. If the button has a pale colour, this means that the corresponding file does not exist. The existing files can be played directly or downloaded individually. The line 'date of recording' contains the date of the recording as well as the age of the speaker on that day (in brackets). The pathologies and the keywords (Remarks w.r.t. diagnosis) that are sought are marked red, as before. If a file with comments on a certain recording session exists, its contents are shown under 'comments'.
After selecting the recording sessions, you can choose on this page which of the files from the recording session are to be exported and in which format. You can choose between:
For each of the listed recordings, the speech and EGG signals are stored. Both signals can be exported in the original file format (NSP and EGG format) as well as in WAV format. If no format is selected for one of the signals, the corresponding signal files are not exported. Click on the button 'Accept' to confirm the selection. This generates a zip file with the selected files on the server. The process may take several minutes. To show the progress of this process, the page will be actualised every ten seconds. If the web-browser does not support automatic forwarding, you can actualise the internet page by reloading manually. In the end, a link appears to download the generated zip file.
Error messages that are made during the generating of the zip file (e.g. a missing file) are saved in a file with the extension '.log'. The file is archived in a zip file as well, if this is possible. Its contents are also shown on the internet page.
To save space in memory, the zip-file is deleted from the server automatically after a some time, depending on the configuration of the server.
You can return to the internet page with the search criteria by clicking on the button 'Back'.