Two-Tier Feature Extraction with Metaheuristics-Based Automated Forensic Speaker Verification Model

While speaker verification represents a critically important application of recognition, it is also the most challenging and least well-understood application. Robust feature extraction plays an integral role in enhancing efficiency forensic verification. Although speech signal continuous one-dimensional time series, recent models depend on recurrent neural network (RNN) or convolutional (CNN) models, which are not able to exhaustively represent human speech, thus opening themselves up forgery. As result, accurately simulate further ensure authenticity, we must establish reliable technique. This research article presents Two-Tier Feature Extraction with Metaheuristics-Based Automated Forensic Speaker Verification (TTFEM-AFSV) model, aims overcome limitations previous models. The TTFEM-AFSV model focuses verifying speakers applications by exploiting average median filtering (AMF) technique discard noise signals. Subsequently, MFCC spectrograms considered as inputs deep network-based Inception v3 Ant Lion Optimizer (ALO) algorithm utilized fine-tune hyperparameters related model. Finally, long short-term memory (LSTM-RNN) mechanism employed classifier for automated recognition. performance validation was tested series experiments. Comparative study revealed significantly improved over approaches.