IDENTIFIKASI HEWAN BERDASARKAN POLA AKUSTIK DENGAN PRINSIP EKSTRAKSI WAVELET DAN KLASIFIKASI MULTI-LABEL JARINGAN SYARAF TIRUAN

Abstract

An acoustic identification and classification system of frogs has been designed based on the principle of wavelet extraction and label classification using an artificial neural network (ANN). This system consists of electronic detection for frog audio as well as an interface using the MATLAB 2018b software as an ANN provider device. As input for the neural network, 5 types of frogs were used, namely the rock frog (Limnonectes macrodon), the blentung frog (Kaloula baleata), the hip frog (Limnonectesblythii), the rice field frog (Fejervarya cancrivora), and the trench frog. frog. frog (Fejervarya limnocharis). ), each with 12 sound samples. Before being inserted into the neural network, 3 levels of sound samples were extracted and denoised using wavelet symlet 3. Furthermore, in the neural network training process, 3 validation samples and 3 test samples were used. After training, the artificial neural network was able to identify the type of frog being tested.