Brandstetter, ChristinaPraher, Katharina2023-10-172023-10-172023https://dspace.htl-perg.ac.at/handle/htl-perg/813CEPP is short for Communicating Emotion by Playing the Piano. The goal of the project is to recognise the intended emotion of someone playing the piano by analysing the generated user input. The project is made up of three main components: Communication between the Computer and the E-Piano via MIDI ensures, that user input can be processed. The application cuts the incoming MIDI-data into small segments, which are then further analysed. A machine learning model that was trained with a public dataset determines an emotion for these segments. Part of the project was comparing and testing different model classes. There are four emotions available, namely Happy, Angry, Sad and Tender. These four were chosen, since the used dataset is split into corresponding quadrants. The exploration and assessment of the existing data prove to be another central part of the project. Especially determining and evaluating different features for the available MIDI-data influences the resulting model greatly. The model classifies the recorded data and provides a category, which is displayed over a graphical user interface. The depiction consists of a colour-coded label and describes the determined emotion. The end result consists of the full application, which is made up of python modules and PowerShell scripts. The optimised model is an important part of this application. Furthermore, additional data was recorded to enhance the existing dataset and train different models.de