Separating Overlapping Birdsongs Enhances the Reliability of Avian Vocal Activity Analysis

Area of Science:

• Bioacoustics
• Ecological monitoring
• Deep learning applications in ecology

Background:

• Bird vocalizations are crucial for ecological research, aiding in population and habitat assessment.
• Passive acoustic monitoring is challenged by overlapping calls and background noise, affecting data reliability.
• Accurate analysis of bird sounds is vital for understanding environmental impacts on avian populations.

Purpose of the Study:

• To develop and validate an improved deep learning algorithm for separating mixed bird vocalizations.
• To assess the impact of call separation on the accuracy of automated species identification and ecological analysis.
• To demonstrate the algorithm's effectiveness in revealing diurnal and seasonal vocalization patterns and environmental influences.

Main Methods:

• An improved deep learning algorithm employing a counter and decoder to process mixed bird calls of uncertain numbers.
• Implementation of three distinct front-end separation strategies to evaluate their effect on classification performance.
• Field validation in the Huangmaohai Cross-Sea Channel area, southern China, using diurnal (White Wagtail) and nocturnal (Black-crowned Night Heron) species.

Main Results:

• Call separation effectively filters noise and interfering species calls, correcting original monitoring biases.
• Accurate diurnal activity patterns were revealed, eliminating false peaks and restoring masked nocturnal rhythms.
• Enhanced statistical sensitivity demonstrated the suppression of bird vocalizations by high temperatures and precipitation, and uncovered seasonal patterns.

Conclusions:

• The developed deep learning algorithm significantly improves the reliability of automated bird sound analysis in complex acoustic environments.
• Call separation is essential for accurate ecological assessments, particularly in multi-species resonance scenarios.
• This method provides robust support for long-term ecological monitoring and understanding environmental effects on bird populations.