An incremental technique for real-time bioacoustic signal segmentation

摘要: An incremental transformation of ZCR and energy without using temporal windows.With our method is possible to save memory transmission costs.Solution process large amounts data by resource-constrained devices as WSN. A bioacoustical animal recognition system composed two parts: (1) the segmenter, responsible for detecting syllables (animal vocalization) in audio; (2) classifier, which determines species/animal whose belong to. In this work, we first present a novel technique automatic segmentation anuran calls real time; then, assess performance whole system. The proposed performs an unsupervised binary classification time series (audio) that incrementally computes exponentially-weighted features (Energy Zero Crossing Rate). proposal, classical sliding windows are replaced with counters give higher weights new data, allowing us distinguish between syllable ambient noise (considered silences). Compared sliding-window approaches, associated cost proposal lower, processing speed higher. Our evaluation component considers three metrics: Matthews Correlation Coefficient point-to-point comparison; WinPR quantify precision boundaries; (3) AEER event-to-event counting. experiments were carried out dataset 896 seven different species anurans. To evaluate system, derived four equations helps understand impact recall has on task. Finally, show segmentation/recognition improvement 37%, while reducing communication. Therefore, results suggest suitable systems, such Wireless Sensor Networks (WSNs).