OPTIMIZING AND ASSESSING THE PERFORMANCE OF AN ALGORITHM THAT CROSS-CORRELATES ACQUIRED ACOUSTIC EMISSIONS FROM INTERNALLY FEEDING LARVAE TO COUNT INFESTED WHEAT KERNELS IN GRAIN SAMPLES

An algorithm was developed with optimizable parameters to match sounds from individual insects in grain by cross-correlating signals from an acoustic sensor array. The algorithm was optimized in a series of tri als conducted in the sample chamber of an Acoustic Location 'Fingerpri nting' Insect Detector (ALFID). The sample chamber was filled with uni nfested wheat, except for a single kernel, which was infested with an immature rice weevil. This kernel was placed at a known location in th e sample chamber. With analysis parameters optimized, the algorithm su ccessfully detected the single insect in 100% of the trials, The algor ithm's capability to count multiple insects was assessed by combining signals in data files collected from single insects into a set that re presented sounds from a pair of insects, In these analyses, the algori thm correctly detected the two insects in 100% of combinations three s ensor spacings apart, 100% of combinations two sensor spacings apart, and 70% of combinations one sensor spacing apart. Based on these resul ts and the dimensions of the ALFID sampling chamber, the algorithm has a 90% probability of identifying two randomly located insects produci ng sounds in a wheat sample. Published by Elsevier Science Ltd.