PAMalyzer: HTW Dresden

Javascript ist deaktiviert!

Bitte aktivieren Sie Javascript in Ihrem Browser.

Fakultät Landbau/Umwelt/Chemie

PAMalyzer - Audio Classification and Analysis Software

Overview

PAMalyzer is the all-in-one solution for biodiversity analysis in bioacoustic recordings.

The free and open source software features

Display of spectrograms and quick navigation through audio files
Manual labelling of sounds in audio files
Classification using different BirdNET-based models (BirdNET Lite, BirdNET v2.4, AnuraPAM, OrthopteraPAM coming soon)
Quick manual review of classification results using audio replay and spectrogram display
based on AviaNZ

The aim of developing PAMalyzer is to provide an easy-to-use software suite that bridges the gap between raw classifier output and verified species detections.

Kontakt

M.Sc. Florian Meerheim

Background

Passive acoustic monitoring (PAM) has emerged as a powerful tool in a variety of disciplines and environments. However, the large volume of data typically collected by this method often exceeds the capacity of humans to analyze it manually. Classification algorithms such as BirdNET can assist with this by automatically identifying specific sounds, such as bird songs and calls, in recordings. Although BirdNET provides a confidence level for each identification, but this level does not accurately reflect the probability of a true identification. Consequently, some of the identified sounds must be reviewed manually to ensure accuracy.
Existing tools that enable users to perform this task mainly focus on research applications. What is missing is a user-friendly tool that makes it easy to quickly select and review recordings that require closer examination. This is one reason why passive acoustic monitoring isn't yet widely used in conservation efforts and environmental planning.

PAMalyzer is a tool that fills this gap by providing a single software package that includes automatic identification, visual analysis of sound files and manual verification.

Setup

For instructions on how to install the software on your system, please refer to the PAMalyzer GitHub page.

Workflow

The general workflow for the analysis of audio data from PAM studies, collected in the field with autonomous recording units (ARUs), comprises three main steps:

The classification of the audio recordings
The manual review process
The compilation of results

The figure shows how this general workflow is implemented within PAMalyzer.

We also provide a comprehensive manual with detailed instructions on how to translate the workflow into specific actions in PAMalyzer. You can find this in the 'Help' menu of the software.

The picture shows the flowchart for the general workflow for the analyis of audio recordings with BirdNET. It comprises 3 main steps, the classification of audio recordings, the manual review of the classification output and the compilation of results. The manual review process is further divided in three smaller steps, namely the selection of candidate recordings, the actual review of the classifier detections and the annotation of recordings. The graphic further shows how these individual steps are implemented in PAMalyzer. — Schematic overview of the implementation of the general workflow for analysing acoustic data derived from passive acoustic monitoring in PAMalyzer.

AnuraPAM: Classifiers for European Anurans

AnuraPAM is a flexible ready-to-use collection of automated call classifier models for the Central European anuran fauna, trained using the deep-learning framework BirdNET. The models can differentiate 14 Central European amphibian species including one hybrid taxon (Pelophylax kl. esculentus). All our models are available in Zenodo as ready-to-use classifiers for the application in PAMalyzer.

For that purpose use the option to specify a custom classifier in the advanced settings area in the BirdNET classification dialog.