Segregação de Voz Usando Mascaramento INM sobre o Banco de Filtros Gammatone

Christian Arcos Gordillo, Marley Vellasco, Abraham Alcaim

DOI: 10.14209/sbrt.2018.61

Evento: XXXVI Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2018)

Keywords: Enhancement mask Neighborhood TimeFrequency noise.

Abstract

This paper presents an innovative approach that employs an ideal neighbourhood mask (INM) that has the ability to efficiently use Local Binary Pattern (LBP) to indicate which Time-Frequency units of the corrupted voice are dominated by noise. Experimental results obtained with a DNN based voice recogniser in noisy environments demonstrate that the proposed technique achieves significant improvements in terms of word error rate corroborating the superiority of the proposed scheme in comparison with the traditional masking algorithms IBM and IRM

Download