Signal processing the acoustics of honeybees (APIS MELLIFERA) to identify the "queenless" state in Hives

Abstract

The honeybee, Apis Mellifera, is of vital importance to the agricultural sector across the World, primarily due to it exceptional abilities to pollinate crops. In 2006, honeybee colony numbers in developed countries experienced a dramatic decline, due to a diverse range of factors collectively known as Colony Collapse Disorder. At about the same time, revenue from using bees for pollination purposes surpassed revenue from honey production for the first time. Hive inspection is time consuming, disruptive and stressful to the colony, and can be the cause of accidental queen death. Furthermore, a colony without a queen will die unless a substitute queen is successfully introduced. A non-invasive method of hive monitoring is a therefore desirable objective. In this paper, we propose and investigate some acoustic-based methods, based on spectrographic analysis and inspired by established techniques commonly used in the analysis of human speech, to distinguish the "queenright" (i.e. with a live queen present) and "queenless" states. Namely, we compare the spectrograms, FFT and S-transform of the audio recordings. In order to assess the different methods, the results of the frequency analysis are classified using a Kohonen Self-Organising Map (SOM) artificial neural network, which is a useful tool for clustering and visualisation of high dimensional data into a lower dimension space. We evaluate our approach using acoustic data recorded from real beehives, under controlled conditions, over the course of a week, kindly provided by Arnia Ltd, a company specialising in beehive monitoring equipment.