Improving biodiversity assessment via unsupervised separation of biological sounds from long-duration recordings

Tzu-Hao Lin,Yu Tsao,Shih-Hua Fang

doi:10.1038/s41598-017-04790-7

Abstract

Investigating the dynamics of biodiversity via passive acoustic monitoring is a challenging task, owing to the difficulty of identifying different animal vocalizations. Several indices have been proposed to measure acoustic complexity and to predict biodiversity. Although these indices perform well under low-noise conditions, they may be biased when environmental and anthropogenic noises are involved. In this paper, we propose a periodicity coded non-negative matrix factorization (PC-NMF) for separating different sound sources from a spectrogram of long-term recordings. The PC-NMF first decomposes a spectrogram into two matrices: spectral basis matrix and encoding matrix. Next, on the basis of the periodicity of the encoding information, the spectral bases belonging to the same source are grouped together. Finally, distinct sources are reconstructed on the basis of the cluster of the basis matrix and the corresponding encoding information, and the noise components are then removed to facilitate more accurate monitoring of biological sounds. Our results show that the PC-NMF precisely enhances biological choruses, effectively suppressing environmental and anthropogenic noises in marine and terrestrial recordings without a need for training data. The results may improve behaviour assessment of calling animals and facilitate the investigation of the interactions between different sound sources within an ecosystem.

Highlights

Evaluating the effects of environmental change and human development on biodiversity is a critical task in conservation management[1, 2]
To evaluate the performance of periodicity coded non-negative matrix factorization (PC-negative matrix factorization (NMF)), we prepared three types of biological sounds (Fig. 1a), one type of transient broadband noise and one type of environmental noise (Fig. 1b), and a mixture of these five sources to use in our simulated long-term spectrogram (Fig. 1c)
The basis matrix represented the spectral dictionary of the input spectrogram, and each sound source appeared to be composed of multiple bases

Summary

Introduction

Evaluating the effects of environmental change and human development on biodiversity is a critical task in conservation management[1, 2]. Instead of trying to identify all calling species, several ecoacoustics indices have been developed to evaluate the biodiversity changes on the basis of acoustic data. Increases when ecoacoustics indices are used to investigate the spatial and temporal changes in biodiversity in noisy environments. The uncertainty of ecoacoustics indices in noisy environments can be decreased if noise can be effectively suppressed before acoustical analysis. The conventional spectral restoration methods can provide satisfactory performance under most conditions, there is still room to further improve them for non-stationary noisy conditions. In field recordings, both environmental and anthropogenic noises are usually non-stationary. The conventional spectral restoration methods may not be sufficient for improving the performances of ecoacoustics indices

Methods

Results

Discussion

Conclusion