The goal of this exploration is to identify secluded speech signal with the aid of DNN. In this scheme, different verses enunciated by diverse people are deliberated as input speech signal|(SS). The properties of these signals are mined by means of Mel Frequency Cepstral Coefficients (MFCC), Zero Crossing Rate (ZCR). The extracted features are then input to the Deep Neural Network (DNN) for training. DNN utilize for predicting the isolated word speech into text. This research work plans to configure optimal weight values for DNN to enhance performance further. To identify optimal weight values through the manual trial-and-error process will take a long time for computation. This research includes optimization techniques to determine optimal weights values to avoid computational complexity. The consequence of the scheme validates that the accurateness of the suggested Hybrid Grey Wolf Optimization-Whale Optimization Algorithm (GWO-WOA) technique is 94.4%, which outperforms the other existing methods.
Volume 12 | Issue 3
Pages: 304-315
DOI: 10.5373/JARDCS/V12I3/20201195