Novel approach of Multiclass Speech emotion classification by density clustering with convolution neural network Features with SVM

Abstract

In recent studies conclude that speech contains rich information about emotion.so using
this information to understand the human interactions. Classification of emotions by speech
information is challenging because analysis the pattern of features will change human to human and
speaker to speaker. The proposed approach use Deep Learning specific architecture CNN
(Convolution Neural Network) and clustering by Dense stream Clustering. The proposed approach
use transfer features approach which uses CNN layers for extracting different features from
different layers because different layers generate dynamic features pattern due to its activation
function. We Use these features in SVM (Support Vector Machine) for learning because it
demonstrate the difference between Na Yang et al. and our approach.