Proposed Model of Speech Recognition using MFCC and DNN

Abstract

Speech Recognition is the ability of the machine to identify the word or phrases in human language and convert it into machine understandable form. Speech Recognition allows you to supply input to an application together with your voice. Speech recognition systems aim to form human machine communication quickly and simply. The main focus of the project would be to convert the speech of a human into text. In this paper, we propose a system architecture that will fetch speech data, process it and give out an effective text outcome. For Demonstration, we use Polygon smoothing algorithm to pre-process the data, and then use MFCC for feature extraction and we did the comparative study of classification techniques like SVM models and DNN. Although the system provides classification using SVM, in comparison, our system proves to be more precise and accurate if we consider big data to be processed. Over the last few years, deep neural networks (DNNs) have become increasingly popular in many areas including ASR, so we have carried out a detailed survey about Speech Recognition using DNN.