The Red Palm Weevil (RPW) is the most destructive pest of the date palm in the world and a serious pest of coconuts. The insect has caused up to 20% loss of these plantations in Asia and the Middle East. The economic damage to palm crops due to RPW could be mitigated significantly by bioacoustics recognition in an earlier phase of infestation and by applying the appropriate treatment. This study is conducted under the hypothesis that distinctive spectral and temporal features in RPW larval sounds can be combined to construct improved indicators for automated detection of infestations. In this paper, a signal processing system is developed with available acoustic technology to detect the existence of RPW in a tree through its feeding sounds. A large set of features are extracted, including unconventional features such as temporal roll-off, temporal slope and temporal spread. Additionally, an analysis is provided of the criteria for the choice of the optimum frame length, as well as the selection of the suitable window function. The results confirm the efficiency of the developed system with the selected representative features, window functions and frame length to detect the existence of the RPW.
Red Palm Weevil, bioacoustics detection, signal processing, feature extraction, feature selection.