Purging of silence for robust speaker identification in colossal database

The aim of this work is to develop an effective speaker recognition system under noisy environments for large data sets. important phases involved in typical identification systems are feature extraction, training and testing. During the extraction phase, speaker-specific information processed based on characteristics voice signal. Effective methods have been proposed silence removal order achieve accurate work. Pitch Pitch-strength parameters extracted as distinct features from input speech spectrum. Multi-linear principle component analysis (MPCA) utilized minimize complexity parameter matrix. Silence using zero crossing rate (ZCR) endpoint detection algorithm (EDA) applied source utterance during phase. These useful later classification where made basis support vector machine (SVM) algorithms. Forward loking schostic (FOLOS) efficient large-scale SVM that has employed among speakers. evaluation findings indicate suggested increase performance amounts noise ecosystems.