摘要：AbstractRecently, more attention is being payed to the ecological environment because of the rapid decline of animal species especially birds. Wireless Multimedia Sensor Networks (WMSN) can be leveraged to monitor, protect and assess changes in birds populations by capturing and sending images in real time. To deal with their limited bandwidth and energy resources, we propose to reduce the amount of visual data to report by sending only images of interest to the collect station. This can be achieved by identifying the endangered target species locally based on their calls before triggering the camera. However, one prominent obstacle is that in wilderness environment, it is difficult to obtain good quality audio signals. This is due to interfering environmental noises that can mask vocalizations of interest so the audio recognition is likely to fail. Therefore, we set up an automated birdsong recognition along with an appropriate noise reduction-based method in order to improve the audio classification accuracy. This ultimately leads to a better use of the network energy and avoids its waste in transmitting useless images.
