SPUR Project

2009 Winner (9 of 13)

Extracting the Biometric Signature found in Honey Bee Hive Vibrations using Principal Component Analysis
Supervisor: Dr Martin Bencsik, Prof Michael Baxter

The aim of this project was to analyse existing digital honey bee vibrational data in order to see whether a useful biometric signature can be discovered. An experiment had been set up in May 2008, in which two accelerometers have been secured to two separate honey bee hives. These transducers pick up the vibration coming from the entire population of honey bees, and a computer sitting nearby digitises and stores the time course of the vibrations. Student tasks included writing a matlab routine to calculate and store the long term and short term power spectrograms of the data set thereby providing 2D images on which to apply the PCA method, application of the PCA on the set of images for an increasing number of images until at least one full day of vibration data was covered, establishing how many components are relevant to the reconstruction process, looking at the time course of the most significant components to see whether they follow a daily pattern, analysis of the biometric relevance of the signature found by PCA, development of software to detect and include new spectrograms whenever necessary and presentation of findings both internally and externally.