Master Thesis/ Project: Target Vocalization Extraction in Zebra Finches

Adapting a Target Speech Extraction (TSE) algorithm for separating overlapping zebra finch vocalizations.

Keywords: Animal communication, vocalization separation, Target Speech Extraction (TSE) algorithm.

 

Research context

Our group works with zebra finches which is a well-studied animal model of human language learning. Our research group has developed a unique 'Birdpark,' providing a naturalistic environment for up to eight zebra finches. This setup enables detailed observation and recording of birds’ behavior and vocalizations. To capture the details of their vocalizations, we employ an innovative approach involving multiple wall microphones and miniature FM-senders attached to each bird. This method allows us to explore the nuances of vocal learning in juvenile birds and understand how social interactions influence this learning process.

  

Available Project

We offer a semester or MSc project that aims to de-mix overlapping vocalizations recorded on microphone channels to obtain the full spectrum of each vocalization of a single bird.

In this master project, you will spearhead the adaptation of an existing Target Speech Extraction (TSE) algorithm to the unique domain of mixed bird vocalizations. Leveraging individual radio signals as guiding cues, we aim to explore the applicability of an algorithm initially designed for human vocal processing to the intricate melodies of avian communication. Additionally, our ambition extends to identifying alternative cues, such as individual vocal signatures, that may enhance the separation performance and contribute to a deeper comprehension of bird vocalizations.

 

Your benefits

You will refine your skills in general machine learning and data analysis. You will collaborate with experts from diverse fields including neuroscience, bioacoustics, behavior, linguistics, and signal processing, gaining a holistic perspective on the intersection of technology and animal communication.

 

Your profile

We are seeking a student with a background and passion for machine learning, audio signal processing, and programming. To apply please send a CV and transcript of records to one of the contacts below.

Zmolikova, K., Delcroix, M., Ochiai, T., Kinoshita, K., Černocký, J., & Yu, D. (2023). Neural Target Speech Extraction: An overview. IEEE Signal Processing Magazine, 40(3), 8-29.

Rüttimann L, Rychen J, Tomka T, et al. Multimodal system for recording individual-level behaviors in songbird groups[J]. bioRxiv, 2022: 2022.09. 23.509166.

 

Collaborators:

Yuhang Wang: yuhang@ini.ethz.ch

Supervisor:

Prof. Dr. Richard Hahnloser: rich@ini.ethz.ch