GAUSSIAN MIXTURE MODEL (GMM) BASED K-MEANS METHOD FOR SPEECH CLUSTERING
K. Rajendra Prasad · International Journal of Advanced Research in Computer Science · 2017
Speech recognition and speaker identification are two important problems in speech clustering. K-means is an efficient clustering method, however it is required to modify this method in speech clustering to the purpose of modeling of speaker voices. In this paper, the pre-processing is performed using HTK and MATLAB Tools. Gaussian Mixture Modeling is used for modeling of speaker voice. In GMM modeling, several parameters are derived for the purpose of defining the shape or structure of a speaker voice. Feature extraction is a process that extracts data from the voice signal that is unique for each speaker. GMM-based k-means method is derived for efficient clustering results and respective results are discussed in experimental section for demonstrating efficiency of proposed method.