Detecting alien invaders

Researchers at York University have developed a system that uses computational bioacoustics to detect alien invasive insect species.

Globalisation of trade is increasing the incidence of invasive pests (such as the Asian Longhorn Beetle) in the UK, and in other countries around the world, often causing significant economic damage.

To monitor these invasive insect species, Dr David Chesmore, a senior lecturer in the Department of Electronics at York University, is developing a system using computational bioacoustics.

Computational bioacoustics works by sensing sounds which are captured using microphones (including ultrasound for bat identification), hydrophones for underwater use, and vibration analysis to identify insects. The results are analysed by computer. The aim is to get a unique ‘fingerprint’ for each sound to allow for identification of a species.

The range of animals that can be analysed using the technique is wide, and includes bats, insects, cetaceans (such as whales, dolphins), amphibians, mammals, birds and aquatic mammals.

One of the aims of Dr Chesmore’s project is to develop a hand-held instrument for use on imported goods. The hand-held device will use vibration sensors to detect the larvae of invasive insects which may enter the country undetected in imported products.

In an exclusive interview with The Engineer Online, Dr Chesmore said that he had already proven that the system works and had secured a Defra contract to build the real-time hand-held and datalogging devices for plant health and seeds inspectors to use.

‘The system I have developed uses low-cost piezoelectric sensors that act as vibration detectors when attached to wood packing, timber or live trees,’ said Dr Chesmore.

Vibrations are caused by insect larvae moving and biting and many of the biting sounds are species specific. Vibrational signals are sampled typically at 44.1kHz and then analysed using time domain signal coding which is a computationally simple method for obtaining features suitable for use in identification.

‘The identification phase is carried out using an artificial neural network, currently a learning vector quantisation (LVQ) type,’ he added. Identification accuracy is currently greater than 95 per cent for three species of insects.

Dr Chesmore’s project is part of a wider, international network which is using the same technology for a range of purposes including monitoring rare and endangered species, conducting rapid biodiversity assessments, and detecting the impact of climate change on species.