AudioMoth
is a low-cost, credit-card-sized acoustic monitoring device used by researchers and conservationists to record wildlife and environmental sounds. Developed by Open Acoustic Devices (a team from the Universities of Oxford and Southampton), it is open-source and capable of recording both audible and ultrasonic frequencies (up to 384kHz), making it ideal for monitoring bats, birds, insects, and even illegal human activities like poaching or logging.
Core Features
Broad Spectrum: Records from 8,000 to 384,000 samples per second, capturing everything from low-frequency frog calls to high-frequency bat echolocation.
Low Power: Powered by 3 AA batteries, it is highly energy-efficient and can be deployed for weeks or even months depending on the recording schedule.
Compact Design: Measuring just 58 x 48 x 15 mm, it is easy to transport and hide in the field.
Storage: Saves uncompressed WAV files directly to a microSD card.
Smart Logging: Can be programmed with custom algorithms to "listen" for specific sounds (like a gunshot) and only record when that event is detected, saving battery and storage.
Variants & Accessories
HydroMoth: A specialized version designed for underwater recording.
MicroMoth: A smaller micro-version for animal-borne monitoring.
AudioMoth Dev: A development board with extra headers for connecting external sensors or GPS.
Enclosures: While the board itself isn't waterproof, official IPX7 waterproof cases are available for terrestrial and underwater use.
Getting Started
To use an AudioMoth, you typically need:
Software: Use the AudioMoth Configuration App to set recording schedules and the Flash App for firmware updates.
Setup: Insert a microSD card and 3 AA batteries. Connect to a computer via USB to sync the onboard clock.
Deployment: Set the physical switch to "Custom" (for your programmed schedule) or "Default" and place it in a protective case.
Analyzing AudioMoth spectrograms for bird identification involves transforming raw audio recordings into visual representations ("fingerprints" of sound) to identify species either through manual visual analysis or automated machine learning classifiers. AudioMoth devices are ideal for this because they capture high-quality .WAV files (typically at 48 kHz for birds) that clearly show frequency and amplitude over time
.
Here is how AudioMoth spectrograms can be analyzed for bird identification:
1. Pre-processing and Setup
Recording Settings: Configure the AudioMoth to 48 kHz sample rate to ensure the full range of bird songs is captured.
File Format: Use uncompressed .WAV files for maximum detail in the spectrogram.
Software: Use software to visualize the audio. Free options include Audacity, Kaleidoscope Pro (viewer), or specialized bioacoustics software like Raven.
2. Manual Analysis (Visual Inspection)
Manual analysis involves looking at the spectrogram as a "heatmap" where the vertical axis is frequency (pitch), the horizontal axis is time, and the intensity (color) represents volume.
Visual Fingerprints: Each species produces a unique pattern. For example, a Cuckoo has a distinct shape, while a wren may look like a dense, complex series of lines.
Identifying Characteristics: Look for:
Line shape: Curves (modulated notes) vs. straight lines (sustained tones).
Temporal pattern: Regular, slow intervals (series) vs. fast, inseparable notes (trills).
Frequency Range: High pitch (top of the graph) vs. low pitch (bottom of the graph).
3. Automated Analysis (Machine Learning/AI)
Given the large amount of data generated, AI tools are the most efficient way to analyze AudioMoth recordings.
BirdNET: This is the most popular tool for analyzing AudioMoth data. It is a Convolutional Neural Network (CNN) that processes 3-second segments of AudioMoth recordings to identify over 6,000 species.
Workflow: Run audio files through the BirdNET-Analyzer (available on GitHub or via a GUI). It provides a species ID and a confidence score (0 to 1).
Performance: Studies show BirdNET can achieve a 72–85% accuracy in detecting bird species, with higher confidence scores reducing false positives.
Other Tools: Kaleidoscope Pro offers cluster analysis to group similar sounds, which is useful for identifying unknown calls.
4. Specialized Analysis (Nocmig)
Nocturnal Migration (Nocmig): AudioMoths are frequently used to record flight calls of birds migrating at night. These calls appear as short, distinct streaks on a spectrogram that can be identified by comparing them to known examples, often using software like Audacity.
5. Data Post-processing
Filtering: Use a "Bird Song Detector" to automatically isolate only the parts of the recording that contain bird song, ignoring background noise. This step has been shown to improve the accuracy of subsequent species identification.
Verification: While AI is effective, expert verification is recommended for rare species or unexpected sounds.
Best Practices for Analysis
Frequency Range: Focus analysis on the 1.5 kHz to 8 kHz range, as this covers most bird species.
Gain Settings: Use "Medium" gain to avoid overdriving the microphone in high-volume environments, ensuring clearer spectrograms.
Context: Check the time and date of the recording, as seasonal and daily timing (e.g., dawn chorus) is crucial for identifying species