DetEdit

DetEdit visualizes and annotates acoustic events from passive acoustic monitoring to evaluate odontocete echolocation click detections and support refinement of automated detectors and classifiers.

Key Features:

• Configurable annotation and classification: Supports annotation and classification of individual signals or clusters and evaluation of signal properties and false detections.
• Comprehensive signal visualizations: Displays time series of received levels, long-term spectral averages, time intervals between detections, scatter plots of peak frequency, root mean square (RMS), and peak-to-peak received levels, and individual or averaged sound pressure waveforms and power spectra per event.
• Event and encounter context: Provides signal-level detail and encounter-level context to facilitate systematic review and quantification across large datasets.
• False detection analysis: Identifies false detections and estimates false positive rates to inform refinement of automated detection methods.
• Generation of labeled datasets: Produces labeled signal datasets for downstream analyses such as training classifiers and quantifying occurrence, abundances, or trends.
• Integration with automated detectors and classifiers: Complements outputs from automated signal detectors and classifiers to enable validation and correction of detections.

Scientific Applications:

• Odontocete echolocation analysis: Evaluation and validation of underwater-recorded odontocete echolocation click detections from passive acoustic monitoring.
• Impulsive-signal bioacoustics: Adaptation for analysis of other stereotyped impulsive signals in bioacoustic datasets.
• Classifier development and ecological inference: Creation of training datasets for classifiers and quantification of occurrence, abundance, or temporal trends from acoustic detections.

Methodology:

Integrates multiple visualization techniques—time series of received levels, long-term spectral averages, inter-detection intervals, scatter plots of peak frequency, RMS, and peak-to-peak received levels, and waveforms and power spectra—to inspect acoustic events, compare with automated detector outputs, produce labeled datasets, and estimate false positive rates.