Identificação automática de aves a partir de áudio

• Autor(a) principal: Carvalho, Silvestre Daniel Dias
• Data de Publicação: 2020
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/16439
• topic: Bird Audio Classification; Deep Learning; Audio Feature Extraction

Bird classification from audio is mainly useful for ornithologists and ecologists. With growing amounts of data, manual bird classification is time-consuming, which makes it a costly method. Birds react quickly to environmental changes, which makes their analysis an important problem in ecology, as analyzing bird behaviour and population trends helps detect other organisms in the environment. A reliable methodology that automatically identifies bird species from audio would be a valuable tool for the experts in the area. The main purpose of this work is to propose a methodology able to identify a bird species by its chirp. There are many techniques that can be used to process the audio data, and to classify the audio data. This thesis explores the deep learning techniques that are being used in this domain, such as using Convolutional Neural Networks and Recurrent Neural Networks to classify the data. Audio problems in deep learning are commonly approached by converting them into images using feature extraction techniques such as Mel Spectrograms and Mel Frequency Cepstral Coefficients. Multiple deep learning and feature extraction combinations are used and compared in this thesis in order to find the most suitable approach to this problem.