A workflow for the automated detection and classification of female gibbon calls from long-term acoustic recordings

• DOI: 10.3389/fevo.2023.1071640
• Publication Date: 2023-02-09T11:36:09Z
• Authors: Dena J. Clink; Isabel Kier; Abdul Hamid Ahmad; Holger Klinck
• Keywords: Terabyte

Passive acoustic monitoring (PAM) allows for the study of vocal animals on temporal and spatial scales difficult to achieve using only human observers. Recent improvements in recording technology, data storage, battery capacity have led increased use PAM. One main obstacles implementing wide-scale PAM programs is lack open-source that efficiently process terabytes sound recordings do not require large amounts training data. Here we describe a workflow detecting, classifying, visualizing female Northern grey gibbon calls Sabah, Malaysia. Our approach detects events band-limited energy summation does binary classification these (gibbon or not) machine learning algorithms (support vector random forest). We then applied an unsupervised (affinity propagation clustering) see if could further differentiate between true false positives number females our dataset. used this address three questions: (1) automated provide reliable estimates patterns calling activity; (2) can approaches be as post-processing step improve performance system; (3) estimate how many individuals (or clusters) there are area? found plateaued with >160 clips each two classes. Using optimized settings, achieved satisfactory (F1 score ~ 80%). The did effectively return clusters appear correspond area. results indicate more work needs done before reliably individual occupying area from Future applying methods across sites different species comparisons deep will crucial future conservation initiatives Southeast Asia.