Research Group at TU Berlin:

Audio Communication

Contact Information at TU Berlin:

Technische Universität Berlin
Faculty I

Sekr. EN 8
Einsteinufer 17
10587 Berlin
Germany

Email: athanasios.lykartsis@tu-berlin.de
Phone: +49 (030) 314 75772

Homepage:
Further Information can be found on http://www.ak.tu-berlin.de/menue/team/wissenschaftliche_mitarbeiter/athanasios_lykartsis/parameter/en/

Research Focus:

The focus of my research lies on developing novel and efficient ways for the extraction of rhythm information from audio signals, whereas the content analyzed might be music or speech. This involves research in the area of rhythm perception, in order to define rhythm as succinctly as possible and detect similarities and differences between music and speech; as well as the development and application of new methods in signal processing in order to design informative audio features for rhythm description. This knowledge can then be used for various tasks such as genre classification (for music) or language identification (for speech), in order to test the validity of the features which represent rhythm. 

In my PhD, I am using rhythm description methods from the field of Music Information Retrieval which can be adapted and used for speech signals. After conducting first experiments on language identification using modifications of the beat histogram method, it has been shown that features in this direction can be informative and useful while avoiding manual annotation. However, there are improvements which can be made to create stronger results and various different approaches to test. At the moment, I have three main areas of activity: 

• The development of a novel rhythm description method for speech, which takes into account higher level speech elements (such as syllables) in an automatic way.
• The further study of (perceptual and other) similarities and differences of rhythm in speech and music, in order to establish rhythmic dimensions.
• The implementation of rhythmic feature extraction methods in music, which can then be further ported to speech signals.

In summary, my research interests include the following areas in Speech Technology and Music Information Retrieval:

• Language and Speaker Identification
• Genre Classification and Recommendations
• Machine Learning for Audio Content Analysis
• Computational Rhythm Description
• Rhythm Perception and Modeling