Secret Sounds of Bees: Analysis of Honey Bee Vibroacoustics Using Hidden Markov Models

Overview

Pollinators around the world are declining at a rate faster than ever recorded due to pesticides, diseases and pests, and habitat loss. Unfortunately, honey bee colony loss is difficult to prevent because early warning systems for colony health are lacking. I developed an early warning honey bee health detecting system that uses a machine learning model and vibroacoustic signals to provide information about the health of a colony before it is lost. Vibroacoustics are sounds and vibrations that are emitted by bees. I developed a Hidden Markov Model within MATLAB using a Hidden Markov Model Toolkit for MATLAB (MATLABHTK). Nine health states were included in the model, and 5-minute recordings were recorded at least weekly from 25 hives in Iowa from August-November, 2021 and processed through the model. The model was 100% accurate in identifying the signals from the training repository and 92% accurate when the entire collection of 258 recordings was assessed. This is the first reported model that provides beekeepers with a non-invasive analysis of their colonies’ health that identifies vital situations like volatile chemical exposure, robbing, active honey flows, etc. This model can be used to reduce colony loss rates when combined with mitigation strategies from beekeepers.

Competition history

• JSHS 2022 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.