Abstract

Over the previous years, a marvelous quantity of study was performed by utilizing the artificial intelligence based deep learning approaches for the speech recognition applications. The automatic speech recognition (ASR) facing the problems in as preprocessing, feature extraction and classification stages mostly, thus solving these problems is mandatory to improve the classification accuracy of speech processing. To solve these issues, an advanced speech recognition methodology has developed by utilizing the Spectral Subtraction (SS) method of denoising with the combination of Mel-frequency Cepstral coefficients (MFCCs) and linear predictive coefficients (LPCs) feature extraction of speech signals. Then back propagated artificial neural networks (BP-ANN) is utilized for classifying the speech signals for the purpose of ASR, respectively. The simulation results show that the proposed approach gives the better classification accuracy compared to the state-of-ASR approaches.