Distinct Vocalization Assignment for On-Bird Syllable Analysis

Among all animal forms for communication, vocal expression contains much information about an individual’s physiological and behavioural states. Thus, to better understand animal’s behaviours and group dynamics, it is of utmost importance to study vocal patterns and their meanings. Although it is possible to record animal vocalizations with external microphones, more insights by a clear allocation of sounds are gained from miniature sensors mounted directly on the animal’s back. Distinguishing and assigning the single vocalizations to the originator is essential for identifying social structures. For this a vocalization recording device with a very selective audio recorder – each device should only record the assigned bird’s vocalizations – is needed.

Among all animal forms for communication, vocal expression contains much information about an individual’s physiological
and behavioural states. Thus, to better understand animal’s behaviours and group dynamics, it is of utmost importance to
study vocal patterns and their meanings. Although it is possible to record animal vocalizations with external microphones,
more insights by a clear allocation of sounds are gained from miniature sensors mounted directly on the animal’s back.
Distinguishing and assigning the single vocalizations to the originator is essential for identifying social structures. For this a
vocalization recording device with a very selective audio recorder – each device should only record the assigned bird’s
vocalizations – is needed.

The main goal of this thesis is to collect audio data with a bone conduction grade accelerometer and compare the recordings with a classical MEMS microphone in terms of noise rejection. For this, an already existing ultra-low power and 1.5 grams heavy audio recording backpack for bird is used. Recorded audio is directly collected on the bird’s back and sent over Bluetooth Low Energy (BLE) to the computer for post-action analysis. **Prerequisites (not all necessary)** - Good Experience in C-programming with microcontrollers (nRF52832, nrf52833) - Experience with wireless communication, ideally BLE - Basic Knowledge of circuit design - Experience in audio processing is beneficial **Type of Work** - 60 % Firmware development - 20 % Experimental and scientific evaluation - 20 % Data analysis and documentation

The main goal of this thesis is to collect audio data with a bone conduction grade accelerometer and compare the recordings
with a classical MEMS microphone in terms of noise rejection. For this, an already existing ultra-low power and 1.5 grams
heavy audio recording backpack for bird is used. Recorded audio is directly collected on the bird’s back and sent over
Bluetooth Low Energy (BLE) to the computer for post-action analysis.

**Prerequisites (not all necessary)**
- Good Experience in C-programming with microcontrollers (nRF52832, nrf52833)
- Experience with wireless communication, ideally BLE
- Basic Knowledge of circuit design
- Experience in audio processing is beneficial

**Type of Work**
- 60 % Firmware development
- 20 % Experimental and scientific evaluation
- 20 % Data analysis and documentation

Lukas Schulthess (lukas.schulthess@pbl.ee.ethz.ch) Dr. Michele Magno (michele.magno@pbl.ee.ethz.ch)

Lukas Schulthess (lukas.schulthess@pbl.ee.ethz.ch)
Dr. Michele Magno (michele.magno@pbl.ee.ethz.ch)