5: CNN Pre-Processing for Automated Identification and Sorting of Vibrational Acoustic Signals

Presenter(s)

Hebron Bekele

Abstract or Description

The overarching goal of the project is to identify the potential correlation between insect vibrational acoustic communication, vegetation and insect diversity in prairies. To deal with the immense amount of sound data analysis required, we developed an innovative toolset that harnesses the power of Convolutional Neural Network (CNN) deep learning to identify insect vibrational signals from large sound files. This approach has only been previously applied to airborne sounds, and vibrational signals pose unique challenges. This approach will significantly reduce analysis time and facilitate the ability to process large vibrational acoustic datasets.