Ultra-Fast Audio Processing For Deep Learning

Betreuer/in:            Dr. Merten Stender           
Dekanat/Institut:   M-14           

E-Mail:   m.stender@tuhh.de

Ultra-fast Audio Processing for Deep Learning

Are you an audiophile programmer with an interest in mechanical engineering for cars? If so, this can be a perfect fit for your project work! We are looking for an independently working student that is eager to accelerate today’s state-of-the-art audio signal processing for automotive brake systems.

Within the field of research on brake system vibrations, large numbers of microphone recordings are generated to capture many different kinds of noises that are generated in the friction interface, such as brake squeal, moan, groan, and many others. The Dynamics Group (M-14) is developing novel Machine Learning methods for categorizing and detecting those different sounds in the audio files. Due to the vast amount of data, the audio signal processing (reading, converting into the frequency domain, feeding into the ML model) is currently the bottleneck for real-time ready software. This project aims at accelerating signal processing by making use of very fast algorithms (C++, Cython, parallel computing) and lean codes. Hence, we are looking for a person that is interested in audio, and who is not afraid of writing some lines of code.

Required knowledge and experience

  • Fundamental understanding of programming concepts
  • Curiosity, highly motivated, independent and precise work ethics
  • Demonstrated coding experience (please indicate some projects that you have been working on)
  • Readiness to deep-dive into audio files

 

Feel free to get in contact and ask specific questions. Optimally, you would consider some student job in our institute after finishing the project.

 

 

 

 

Schreibe einen Kommentar

Deine E-Mail-Adresse wird nicht veröffentlicht. Erforderliche Felder sind mit * markiert.