DetEdit: A graphical user interface for annotating and editing events detected in long-term acoustic monitoring data

Abstract

Passive acoustic monitoring has become an important data collection method, yielding massive datasets replete with biological, environmental and anthropogenic information. Automated signal detectors and classifiers are needed to identify events within these datasets, such as the presence of species-specific sounds or anthropogenic noise. These automated methods, however, are rarely a complete substitute for expert analyst review. The ability to visualize and annotate acoustic events efficiently can enhance scientific insights from large, previously intractable datasets. A MATLAB-based graphical user interface, called DetEdit, was developed to accelerate the editing and annotating of automated detections from extensive acoustic datasets. This tool is highly-configurable and multipurpose, with uses ranging from annotation and classification of individual signals or signal-clusters and evaluation of signal properties, to identification of false detections and false positive rate estimation. DetEdit allows users to step through acoustic events, displaying a range of signal features, including time series of received levels, long-term spectral averages, time intervals between detections, and scatter plots of peak frequency, RMS, and peak-to-peak received levels. Additionally, it displays either individual, or averaged sound pressure waveforms, and power spectra within each acoustic event. These views simultaneously provide analysts with signal-level detail and encounter-level context. DetEdit creates datasets of signal labels for further analyses, such as training classifiers and quantifying occurrence, abundances, or trends. Although designed for evaluating underwater-recorded odontocete echolocation click detections, DetEdit can be adapted to almost any stereotyped impulsive signal. Our software package complements available tools for the bioacoustic community and is provided open source at https://github.com/MarineBioAcousticsRC/DetEdit.

Cetacean distribution models based on visual and passive acoustic data

“…Echolocation clicks for each target species were detected and classified in the PAM recordings 19 , 31 , 32 . Detections and classifications were manually reviewed by expert analysts to ensure low false positive and misclassification rates using DetEdit a custom software and graphical user interface package written in MATLAB 33 .…”

Technological Developments Since the Deepwater Horizon Oil Spill

“…Data generated by the automated network-based classification technique were used by Solsona-Berga et al (2020) in the development of the DetEdit graphical user interface that accelerates the editing and annotation of automated detections from marine mammal acoustic data sets. K. Li et al (2020) developed a threestage automatic hybrid classifier algorithm that combines a traditional technique with unsupervised clustering to detect species-specific beaked whale calls in acoustic data.…”

A machine learning pipeline for classification of cetacean echolocation clicks in large underwater acoustic datasets

“…Expert analysts have been relied upon to detect and classify events of interest in large acoustic datasets since the 1970s [ 5 ], typically by visually scanning through spectrograms and noting times of individual signals of interest or sets of co-occurring detections (often termed events, acoustic encounters, or bouts). Manual analysis is flexible, effective, and remains a key aspect of many recent passive acoustic monitoring studies [ 6 – 10 ]. Analysts are trained to recognize one or more signal categories of interest, and over time they gain an understanding of the within type variability of these signals, recognize related variants, and learn to minimize confusion related to environmental sounds and non-target signals [ 11 ].…”

“…However, due to the short duration and minimal information contained in each detection, it is nearly impossible for analysts to make consistent detection-level determinations as to whether or not a signal has been correctly labeled. Contextual information, such cue rate and proximity of similar signals, is typically relied upon to make manual classification decisions [ 6 ], but context is lost when data are reduced to a set of independent automated detections. Conversely, too much reliance on context when designing automated classifiers can cause rarer signal types to be overlooked in cases where multiple signal types occur simultaneously.…”