Previous Page 
|
Previous Page |
Next Page |
Annotations are manually entered notes about individual vocalizations found in recordings and contain a hierarchy of classification information in addition to arbitrary comments.
Song Scope defines a hierarchy of classification information included in annotations as follows:
A class is a label used to identify a vocalization at a high level. Typically, this could be the name of a particular species under investigation.
A subclass is a label used to divide vocalizations in a particular class down into meaningful groups for classification. This is intended to divide vocalizations into types that can be compared to each other. For example, many species of birds have a number of calls and songs among their repertoire of vocalizations. The subclass could be used to divide vocalizations into flight calls, contact calls, and spring songs, for example.
The id is a label used to divide vocalizations in a particular class down into different variations. For example, this could indicate different individuals, or it could indicate typical regional variations. Song Scope uses the id to measure the quality of different recognizer models as follows: By building a recognizer model with all the vocalizations excluding a particular id, the ability of the recognizer to detect the excluded vocalization is an indication of how well the model performs. For more information see Pattern Recognition Basics. If you do not specify an id, Song Scope will automatically assign a unique id associated with the recording. In other words, by default, Song Scope will assume that vocalizations found in different recordings are likely to be from different individuals while vocalizations on the same recording with the same class are likely to be from the same individual.
We strongly recommend that you pay attention to the Signal Detection settings and view when making annotations. Annotations should include some of the "quiet" portions immediately preceeding and following the vocalization (in other words, better to crop vocalizations too wide rather than too narrow).
To create a new annotation, select the vocalization to be annotated. Then use the pop-up menu to select the Annotate... command. A dialog will appear in which you can type in the class, subclass, id, and any additional comments you like. The dialog also contains a shortcut menu tree listing all the previously assigned classes, subclasses and ids. You can click on the name of a class, subclass, or id to have fields filled in automatically instead of typing them again each time. Once created, the annotation will appear in the waveform and spectrogram plots as a solid box with the class, subclass, id, and comments displayed near the box.
For multi-channel audio recordings, the mixer per-channel gain and delay settings are saved with the annotation. In other words, if beam forming is used to focus a microphone array in the direction of the vocalization, the mixing parameters are stored with the annotation. The settings for a particular annotation can be restored by left-clicking on an annotation or by selecting the annotation using the forward and backward annotation navigation controls.
To select an annotation, simply right click (MacOS users can use control click instead) in the annotation on either the spectrogram or waveform plots to display a pop-up menu with the following options:
Listen to the annotation through the computer's sound hardware. Notice that you can also simply press the space bar to play the current selection.
This option is only available when working with multi-channel audio recordings. Automatically enables multi-channel mixing and adjusts per-channel delay controls to "focus" the microphone array using beam forming. If you want to update the mixing parameters of an annotation, first set the desired mixing parameters (e.g. with this "Focus Beam" choice), and then edit the annotation.
Automatically adjust the brightness and contrast display controls such that the strongest frequency component in the selection is shown with the "brightest" color and the weakest frequency component in the selection is shown with the "darkest" color.
Automatically adjust the horizontal (time axis) zoom such that the annotation completely fills the waveform and/or spectrogram plots.
Pops up a dialog window so that you can edit the annotation. The current mixing parameters will be applied.
Delete the selected annotation
Save the selected annotation to a .wav file
Song Scope annotation files are designated with the .ssn suffix and are stored in a subdirectory called SongScopeNotes located in the directory containing the original .wav, .aif, or .wac audio file. The .ssn file is given the same base filename as the corresponding audio file. Song Scope note files contain both the annotation information as well as a copy of the actual audio samples that make up the corresponding vocalization (with any mixing parameters applied as described above). This lets you retain the portions of audio that were interesting enough to annotate without the need to retain the much larger original field recordings.
Note that annotation files only store the mixed audio samples and the class, subclass, id and comments labels. None of the display, spectrogram, or detector parameters are saved in the annotation file (they are instead saved to and loaded from recognizer files).
When a .wav, .aif or .wac recording file is opened, Song Scope looks for a corresponding .ssn file in the SongScopeNotes subdirectory and automatically loads the previously saved annotations.
From the "File" menu, select "Save Annotations" to save annotations and their corresponding audio samples to a Song Scope .ssn file.
If annotations were created or edited without being saved and the Song Scope window is closed, Song Scope will ask if the annotations should be saved first.
Song Scope can also open a .ssn file directly to review and playback saved annotations. The annotations are displayed in their original time positions on the spectrogram and waveform plots, with all the empty space between notations filled with background spectrum information.
|
Previous Page |
Next Page |