Acoustics, Analysis & AI (3AI)

The Acoustics, Analysis & AI (3AI) group at the IT:U Austria in Linz explores how mathematics, acoustics, and artificial intelligence can work together. We study interpretable, hybrid AI systems for sound analysis—combining deep learning with mathematical structure and human insight. Our research ranges from signal processing and hearing models to bioacoustics and mathematical foundations, aiming to bridge human and machine understanding of acoustic phenomena.

Connecting mathematics, acoustics, and artificial intelligence 

Nowadays, artificial intelligence is everywhere — impacting our daily lives and transforming nearly all research domains, including acoustics. From speech recognition and noise reduction to spatial hearing and bioacoustics, AI-driven models are changing how we understand, simulate, and process sound. Deep neural networks and other learning-based models have shown enormous potential to benefit society and have become powerful tools for scientific discovery. At the same time, concerns remain that AI may replace or overshadow human reasoning and perception. The grand challenge for the future is therefore to understand how human and artificial intelligence can cooperate and complement each other—especially in the analysis and understanding of sound. 

In the Acoustics, Analysis & AI (3AI) group at IT:U Linz (Prof. Peter Balazs), this interaction is explored scientifically at the intersection of acoustics, mathematics, and machine learning. Our goal is to build rigorous mathematical foundations and develop interpretable, hybrid AI systems that serve human understanding—particularly in the study of auditory phenomena and the design of intelligent acoustic technologies. 

This group cooperates closely with the Acoustics Research Institute of the Austrian Academy of Sciences.  

The group’s research combines theoretical mathematics, signal processing, numerical acoustics, and modern AI, for example in the following complementary project lines: 

• Hybrid Data-Processing: 
  We study how human expert knowledge and data-driven machine learning can be combined effectively. In audio processing, this concerns whether to pre-process input signals using human-designed time-frequency representations or to use fully end-to-end models. Building our recent work on hybrid filter banks, we pursue new mathematical and computational approaches that merge the strengths of both paradigms. 

• Hearing for and with AI: 
  We develop interpretable models of human hearing that use AI to advance psychoacoustic understanding. In the opposite direction, we integrate insights from auditory perception to enable machines to “hear” and analyze sounds more effectively bridging cognitive science and computational acoustics. 

• Bioacoustics with AI: 
  Animal vocalizations and natural soundscapes contain valuable information about ecosystems and behavior. Using deep learning and signal analysis, we study the acoustic communication of species such as mice and elephants and develop AI-based measures for biodiversity and environmental monitoring. 

• AI for Maths – Maths for AI: 
  This bidirectional line connects mathematical analysis and machine learning. We investigate how mathematical tools can help explain and predict the behavior of AI models—focusing on understanding structure, not merely approximations. Conversely, we explore how AI can assist in mathematical reasoning and support human learning in mathematics. 

The 3AI group thus builds bridges between mathematical theory and real-world acoustics, and between human and artificial intelligence. Students and researchers joining the group can expect an interdisciplinary and open research environment that values analytical precision, creativity, and a collaborative spirit across the boundaries of mathematics, acoustics, and AI.