Under Analysis, you will need to decide between the manual or automatic writer identification analysis, and afterwards, a corresponding pop-up window will be opened.
In the newly opened window, one will find a list of features, which represents the highest levels of the feature base (Feature Base), like a tree hierarchy. One can unfold each feature until the end, where one can choose the corresponding feature values for the specific feature.
In order to analyze an unknown handwritten music score, one selects as many features as possible. For each feature, one has to follow the path in the tree hierarchy until a set of example pictures is shown. One selects the most similar picture among them according to the feature on the music score. It will take some time to choose all features but one will get a better result.
The following graphic demonstrates the available pictures of an eighth note with an upward notated stem.
After choosing a certain value, the corresponding feature will be marked by a check mark. One can also choose several values. By pushing the button Identifizieren, the selected set of feature values representing a feature vector will be sent to the server to identify the unknown writer. The system returns a list of possible writers as shown in the following graphic.
In the returning result list, a distance value and an ND-value are shown for each writer in the list. The ND-value stands for the percentage of features, which are used for the distance calculation. The remaining features were considered unknown and taken as NULL values. Since NULL values have influence on the calculated result, the result set is more representative and reliable, the higher the ND-value is. The distance illustrates the similarity of the handwritings. The lower the distance, the more similar the unknown handwriting and the handwriting of the corresponding writer are.
As already mentioned, a check mark in the tree hierarchy is shown when a value was chosen for a feature. In the event the user made a mistake and wants to delete the selected feature value, the user must go to the feature check mark. A small window will appear in which one can click on the link Löschen in order to delete the feature value.
If one would like to delete all feature values of its analysis, then one can push the button Löschen.
The text field for loading feature vectors has been created only for experts. Each value of a feature will be represented by a numerical code, which can be inserted in the text field. A set of feature values is separated by commas.
As shown on the right, one can add a new handwritten music score to the database. This should only be allowed for experts. After selecting the
feature values for the new handwritten music score, one enters the signature and the writer in the text field and sends them to the server that stores the
data in the database.