Avaliação da Variação Total de Espectrogramas para a Extração Cega de Fontes
Giulio G R Suzumura, Ricardo Suyama

DOI: 10.14209/sbrt.2019.1570559004
Evento: XXXVII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2019)
Classical approaches to source separation such as those based on Independent Component Analysis and Sparse Component Analysis are widespread and, depending on the application, yield good results. However, in audio signal separation study, exploring features in other domains may yield better results. Turning audio signals into spectrograms causes them to be interpreted as images, therefore using imaging techniques can bring new perspectives to the problem. In this sense, in this work, based on metrics associated with image sharpness, we evaluated the use of the Total Variation of the spectrogram in the problem of Blind Source Separation, and the preliminary results indicate that the tool can be useful for the construction of a new approach for signal recovery.