Module 39-M-Inf-ADS Auditory Data Science

Non-official translation of the module descriptions. Only the German version is legally binding.

The goal is to give an overview and transfer knowledge of foundations and interrelations in the area of auditory displays, including key definitions, taxonomy, organization system and the conceptual and practical implementation of different sonification methods. Participants will learn criteria for the selection of sonification methods for given problems as well es guidelines for the parameterization of sonification methods. The scientific analytic listening competence will be developed and trained with help of many sonification examples, with focus on the discrimination of structures in sound. Furthermore participants will develop practical competences in sonification, including programming skills in the sound programming language SuperCollider and interactive coding in the NumPy/SciPy ecosystem, with visualizations and method development in python.

The lecture introduces into the research field of Auditory Data Science which connects Data Mining to Audio Technology via methods of Sonification, i.e., ‘auditory data representations’. Here, sound is the medium for exploratory data analysis of complex data, for monitoring changes in complex data streams, for discriminating complex time series that are difficult to understand from visualization alone, for supporting exploratory interactions with complex data, to enhance our insight into data by teaming up visualization and sonification towards multimodal data exploration or perceptualization. Sonification, the systematic, reproducible representation of data as sound is the key to auditory data
science and thus is first comprehensively introduced and defined. In turn, sonification techniques will be presented and discussed: Earcons, Auditory Icons, Audification, Parameter-Mapping Sonification, Model-based Sonification and Wave Space Sonification. Furthermore, selected techniques for inductive data mining, such as estimation of intrinsic dimensionality, topology or clustering of high-dimensional data will be introduced in order to use them within sonification processes. At hand of selected real-world examples for auditory data analysis we will discuss the potential of auditory data science, e.g. sonification of medical data for diagnosis, monitoring and therapy.

Useful but not mandatory: Knowledge of data mining
Cross-references to: Introduction to Sound Synthesis, Ambient Interfaces

—