Classification of Recurrence Plots of Voice Signals Using Convolutional Neural Networks

Luana Rodrigues Barros, Gabriel Gutierrez P. Soares, Suzete Correia, Gabriel Duarte, Silvana Cunha Costa

DOI: 10.14209/SBRT.2020.1570661665

Evento: XXXVIII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2020)

Keywords: convolutional neural networks texture recurrence plots voice signals

Abstract

In the last decade, texture analysis has been widely applied to image classification. This method has great relevance on recognition of patterns in medical images surfaces. Alternatively, many studies have investigated the usage of Convolutional Neural Networks (CNNs) as a technique for classifying texture images. In this study, a low-complexity CNN was applied to recurrence plots of voice signals to distinguish the presence of laryngeal pathology. A data augmentation technique was employed to increase the number of samples. A study using the same dataset and another classification approach was considered for results comparison. The CNN proposed was proven to be more robust with a 12% increase on accuracy compared to the previous work.

Download