Detecção de patologias laríngeas com base na análise dinâmica de sinais de voz utilizando redes neurais profundas

This paper deals with the application of classifiers based on deep neural networks (DNNs) in the discrimination between healthy voice signals and affected by laryngeal pathologies edema, carcinoma, leukoplakia, polyps and vocal fold paralysis. Each voice signal was partitioned into segments, which are inserted in an DNN of 05 hidden layers with 200 neurons and one neuron in the output layer. For the classification a methodology was used that analyzes the dynamic behavior of the segments used for the test. The proposed method provided an accuracy of 85,55±4,39%