Speech Recognition Using Recurrent Neural Network

Abstract

Speech recognition (SR) and understanding have been the subject of years of research. The vocalized method of human communication is called speech. A restricted collection of vowel and consonant speech sound units are phonetically combined to form each spoken word. SR is the process by which a computer or software recognizes phrases and words in spoken language and converts them into a computer-readable format. Modern technology has progressed to interpret meaning through the application of its own set of grammatical rules, even if it still uses continuous dictation. In this study, we suggest using the Recurrent Neural Network technique and the Mel Frequency Cepstral Coefficient (MFCC) for Speech Recognition (SR) to create an alphabetical list of words from the CORPUS database.