Semantic Audio Processing

Humans structure and categorize perceived acoustic signals into hierarchies of meaningful units or auditory objects. In the case of music signals, such auditory objects may correspond to a melodic phrase, a single note, a rhythmic pattern, or the sound of a specific instrument. A general goal of Semantic Audio Processing is to develop methods for analyzing and manipulating acoustic signals in an intuitive way, i.e. in the way humans perceive and understand them. The AudioLabs contribute to this line of research in various ways.

The processing of music-related audio material has become a major strand of research bridging the gap to various disciplines ranging from information retrieval to music sciences. Despite their enormous diversity, music signals possess specific acoustic and structural characteristics that distinguish them from spoken language or other non-musical signals. The AudioLabs research group headed by Prof. Dr. Meinard Müller develops informed processing techniques that specifically address musical properties related to melody, harmony, rhythm, and timbre.

"Semantic Audio Processing" department at the Audio and Multimedia division of Fraunhofer IIS develops methods for decomposing acoustic signals into semantic entities and for handling, modifying and interacting with these audio objects. This opens up new applications in the area of semantic audio analysis and rendering for flexible, high-quality reproduction of audio content over loudspeakers and headphones, including new forms of semantic user interfaces. Another application area for Semantic Audio Processing is object-oriented audio coding for various scenarios, including introducing semantic audio processing concepts into next generation broadcasting standards