Application of EEMD and Hilbert marginal spectrum in speech emotion feature extraction

Ensemble empirical mode decomposition (EEMD) effectively solves the mode mixing problem, which is caused by traditional empirical mode decomposition (EMD). This paper introduces EEMD and Hilbert marginal into nonlinear and unstable speech signal processing, and proposes band energy recognizing emotional speech based on masking effect and Hilbert marginal. The experiments prove that the ability of EEMD to fight aliasing can effectively extract the intrinsic mode of emotional speech; speech emotion feature based on EEMD can reflect emotional information better.