Hybrid Moth Flame Optimization and Teachers Learning based Optimization Algorithm based Artificial Neural Network Model for Speaker Diarization

Speaker diarization or indexing is the procedure of automatically dividing conversation including numerous speakers into homogeneous segments and clustering together all segments, which symbolize a similar speaker. Till now, numerous researches are carried out in this field, but the requirement of precise partitioning procedure gates lagged in particular conditions. However, this work aspires to develop a novel speaker diarization or indexing technique (Telugu language) which at first includes the Mel Frequency Cepstral coefficient (MFCC) based feature extraction. Then, for the process of clustering, a novel optimized Artificial Neural Network (ANN) is developed. The main aim of the process of clustering is to train the ANN and it is performed using the optimization algorithm which updates the ANN weight by exploiting a hybrid Moth Flame Optimization (MFO) and Teaching learning-based Optimization Algorithm (TLBO). Thus, the adopted technique is called Hybrid MFO-TLBO and the overall performance is evaluated with the conventional models regarding diverse performance metrics.