Variações da Decomposição Empírica de Modos para Realce de Sinais de Voz

The empirical mode decomposition (EMD) has been adopted for the suppression of acoustic noise from speech si gnal in the time domain. This paper investigates two alternative methods to decompose and enhance speech signals collected i n noisy environments. After the decomposition, the Hurst exp onent is used to select the most corrupted modes. The results obtai ned in speech enhancement experiments prove that when compared to the EMD the alternative decomposition methods can lead to interesting speech quality and intelligibility improvement