Hierarchical Organization of Human and Animal Vocalizations

Figure 1 a) Different segmentation strategies for animal and human vocalizations. Upper: human vocalizations are segmented based on the semantic units; bottom: zebra finches’ vocalizations are segmented based on the respiratory cycles. b) Identification of Sub-Syllable Elements Across Vocalization Types: Red Frames highlight an identical vocal element that recurs in two different types of song syllables in zebra finches; Orange Frames display a vocal element shared between a call type and a song syllable type

Keywords: animal communication, acoustic signal processing, comparative bioacoustics

Research Context

Songbirds have been utilized as model animals for studying the acquisition and development of human language for decades. The song of a songbird typically comprises repeated sequences of notes, termed phrases, which vary in pitch, duration, and complexity.

To understand the structure and features of birdsong, it is essential to precisely define the basic vocal unit for analysis. Spectral or amplitude-based segmentation strategy has been a cornerstone of avian vocalization analysis for decades, these methods are essentially based on the intervals created by respiratory cycles to distinguish between vocal and non-vocal elements. In contrast, studies of human language often segment vocalizations based on semantic content or syllabic structures. A single, spectrally continuous element in human speech may be divided into multiple discrete vocalization segments. The applicability of similar segmentation approaches to songbirds remains an open question.

Project Description

Our research group focuses on the study of zebra finches, a species of songbird that is famous for their learned, stereotyped songs. These birds are known for repeating a single learned song throughout their lives. In our observations of isolated adult zebra finches, we have discovered that zebra finches reuse sub-syllable level vocal components to construct their entire vocal repertoire. As illustrated in figures, identical vocal elements are recurrently employed across different vocalization types. This reusage of vocal elements is evident in both calls and songs, suggesting the presence of phoneme-like components that function as a foundational layer within the hierarchical structure of their vocalizations.

The aim of this master's project is to apply the newly developed algorithm to analyze human vocal data, drawing parallels and contrasts with the hierarchical organization observed in zebra finch vocalizations. This comparative analysis will help elucidate the structural similarities and distinctions between bird and human vocal communications. The project seeks to establish a novel framework for the detailed analysis of vocalization features and their development, ultimately enhancing our understanding of the underlying principles governing vocal communication across species.

Your benefits

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

Your profile

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

Collaborator: yuhang@ini.ethz.ch

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