Emotional speaker recognition based on i-vector space model

CASA-based speaker identification using cascaded GMM-CNN classifier in noisy and emotional talking conditions

“…The first step was the preprocessing of the audio signal corrupted by noise and room reverberations using binary time-frequency (T-F) masking algorithm, using a CASA approach, via a deep neural network classifier. Mansour et al [17] employed the i-vector approach along with the Support Vector Machine (SVM) classifier as an attempt to boost and enhance the deteriorated performance of speaker recognition under emotional auditory environments. Results showed that the i-vector algorithm resolves the problem of training algorithm complexity that the SVM model suffers from and shows promising results in increasing speaker recognition performance in an emotional context.…”