Improving Indonesian multietnics speaker recognition using pitch shifting data augmentation

Speaker recognition to recognize multiethnic speakers is an interesting research topic. Various studies involving many ethnicities require the right approach achieve optimal model performance. The deep learning has been used in speaker classes high accuracy results with promising results. However, multi-class and imbalanced datasets are still obstacles encountered various using method which cause overfitting decreased accuracy. Data augmentation overcoming problem of small amounts data multiclass problems. This can improve quality according applied. study proposes a pitch shifting neural network called (PSDA-DNN) identify Indonesian speakers. that done prove PSDA-DNN best multi-ethnic where reaches 99.27% precision, recall, F1 score 97.60%.