Bio-Applied Systems Engineering

Bioacoustic Recognition of Grasshoppers (Orthoptera) in the Field

This is an ongoing project to develop an automated grasshopper identification system as an identification aid and for biodiversity studies. The approach employed uses Time Domain Signal Coding (TDSC) and a multilayer perceptron (MLP) artificial neural network (ANN) to recognise individual species.

However, in the UK, there are many interfering sounds such as vehicles, aircraft and farm animals so the system has been trained to recognise these sounds as well to give better discrimination. Previous research concentrated on species recognition in low noise conditions and 95%+ recognition accuracy has been obtained for 25 species of British Orthoptera. Figure 1 shows a block diagram of the recognition system

Field recordings of 4 grasshopper species (Meadow Grasshopper - Chorthippus parallelus (Zetterstedt, 1821), Common Green Grasshopper - Omocestus viridulus (Linnaeus, 1758), Mottled Grasshopper - Myrmeleotettix maculatus (Thunberg, 1815) and Lesser Marsh Grasshopper - Chorthippus albomarginatus (De Geer, 1773)) were made in East Yorkshire during 2002 with Dr Ohya from the Biodiversity Research Group at the Forest and Forest Products Research Institute (FFPRI) in Morioka, Japan. These were used to train a 13 output MLP for the 4 species, 2 car sounds, 4 bird sounds, 1 blow fly sound, 1 aircraft sound and 1 background sound.

Test results show that it is possible to recognise the species with 80-100% accuracy under field conditions. In addition, it is possible to simply divide the sound into (typically) 2 second blocks and recognise the sound without the requirement for detecting the onset of the sounds.

Figure 2 shows recognition of Omocestus viridulus (labelled OV) in an 18s sequence on a 2 second interval recorded at Allerthorpe Common (near York) on 15 July 2002. The system correctly recognises 3 short songs by O. viridulus, a light aircraft (Plane) and a bird alarm call (Bird1).

Similar work is under way on Japanese cicadas – Tibicen bihamatus, T. flammatus and T. japonicus in Iwate Prefecture in Japan for habitat studies. A hand-held instrument is being designed to automatically recognise these species which will be field tested in 2004 by Dr Ohya (FFPRI).

Figure 1 - Schematic diagram of identification system. The input signal is coded to give (D,S) pairs which are converted into a code per epoch C(n). Codes and lagged codes are accumulated for the duration of the complete sound and the resulting A-matrix forms the input to an artificial neural network. The result is a species identification per sound.

Figure 2 - Classified sounds from an 18s sequence on a 2 second interval recorded at Allerthorpe Common on 15 July 2002. The system correctly recognises 3 short songs by O. viridulus (OV), a light aircraft (Plane) and a bird alarm call (Bird1).

References Related to Automated Identification
Chesmore, E.D., Swarbrick, M.D. & Femminella, O.P. (1998) Automated analysis of insect sounds using TESPAR and expert systems - a new method for species identification. In: Information Technology, Plant Pathology and Biodiversity (Ed: Bridge, P., Morse, D.R.; Scott, P.R.), CAB International, Wallingford, UK, 273-287.

Chesmore, E.D. & Nellenbach, C. (2001) Acoustic Methods for the Automated Detection and Identification of Insects, Acta Horticulturae, 562, 223-231.

Chesmore, E.D. (2001) Alication of time domain signal coding and artificial neural networks to passive acoustical identification of animals, Journal of Alied Acoustics, 62, 1359-1374.

Chesmore, E.D. & Ohya, E. (2004) Automated identification of field-recorded songs of four British grasshoers using bioacoustic signal recognition. Bulletin of Entomological Research, 94(4), 319-330.

Chesmore, E.D. (2004) Automated bioacoustic identification of species. Anais da Academia Brasileira de Ciencias, 76(2), 435-440.

Gardiner, T., Hill, J., & Chesmore, D. (2005) Review of the methods frequently used to estimate the abundance of Orthoptera in grassland ecosystems. Journal of Insect Conservation, 9(3), 151-174.

Other Orthoptera-related References
Chesmore, D. (2005) Records of the Slender Ground-hoer Tetrix subulata (Linn. 1758) (Orthoptera: Tetrigidae) in Yorkshire. Naturalist, 130, 59-64.

Copies available on Request