ABSTRACT
ATLAS is a graphical tool for the annotation of multi-modal data streams. In our application, the data is collected in human computer interaction (HCI) scenarios. Basically, any type of input data can be processed by the ATLAS annotation tool, including speech (multi-channel), video (multi-channel), EEG, EMG, ECG data. In addition to the raw data, intermediate data processing results such as extracted features, and even (probabilistic or crisp) outputs of pre-trained classifier modules can be displayed. Tools for the annotation and transcription of HCI scenes are integrated as well. In this paper ATLAS's basic architectures and features are described, furthermore it is explained how different types of data and label information can be presented. Besides these basic annotation features, an active learning module (active data selection) has been integrated . Currently, probabilistic Support Vector Machines (SVM) are available in the ATLAS system.
