Multi-Channel Time-Domain Boring-Vibration-Enhancement Method Using RNN Networks.

Area of Science:

• Agricultural Entomology
• Acoustic Signal Processing
• Pest Management

Background:

• Wood-boring beetle larvae infest tree interiors, causing damage with minimal early external signs.
• Acoustic monitoring detects larval feeding vibrations, outperforming traditional methods but susceptible to environmental noise.

Purpose of the Study:

• To develop an advanced acoustic signal separation system for accurate detection of wood-boring beetle larvae.
• To mitigate the impact of environmental noise on acoustic detection of wood-boring pests.

Main Methods:

• A multi-channel attention mechanism system was designed to process acoustic signals from multiple sensors.
• Attention mechanisms were used to weight relevant sensor channels, enhancing signal-to-noise ratio.
• Beamforming techniques were applied to effectively remove environmental noise and isolate larval vibration signals.

Main Results:

• The developed system demonstrated efficient separation of wood-boring beetle vibration signals from mixed noisy signals.
• Experimental validation using field and laboratory data confirmed the system's effectiveness.
• The multi-channel attention mechanism proved crucial in noise reduction and signal isolation.

Conclusions:

• The proposed signal separation system significantly improves the accuracy of acoustic detection for wood-boring beetle larvae.
• This technology offers a robust solution for early pest detection in forestry and agriculture.
• Further research can explore optimizing the attention mechanism for diverse noise environments.