Speaker Recognition Using MFCC &Evolution with Different Classification Techniques

Journal Title: International Journal of Engineering and Science Invention - Year 2017, Vol 6, Issue 9

Abstract

This research paper is designed a unimodal biometric system based on speaker recognition. TheMel-Frequency Cepstral Coefficients (MFCC)is used for extracting the voice features. We applied this technique for identification of a person on KVKRG voice database. For this experiment, 500 phonemes of 50 subjects from KVKRG voice database were used. The database consists of phoneme in language spoken in multimodal biometrics research lab, Dept. of Computer Science and Information Technology, Dr. Babasaheb Ambedkar Marathwada University Aurangabad of Maharashtra. The phoneme of each subject is collected five times i.e. English alphabets (a – z) and five times i.e. digits (0 – 10).Three samples used for training and the rest two used for testing. Different classifiers were used like Linear Discriminate, SVM, k-NN, Ensemble Subspace Discriminate. The recognition rate for (a-z) is 98% and (0-10) is 96%. Experimental result shows that biometrics system gives improvement in the overall system performance, results quickly and accurate.

Authors and Affiliations

Pravin G. Sarpate, Ramesh R. Manza

Keywords

Related Articles

Effect of Al2O3 – 40 wt. % TiO2 ceramic particle size and shape on flexural properties of glass fibre reinforced epoxy composites

The objective of this study is to investigate the flexural properties of glass fibre reinforced epoxy composite filled with different form (fused and crushed, agglomerated), size and proportions of Al2O3 – 40 wt. % TiO2...

Empirical Design Equations For PVC-Concrete Composite Columns

In this research, an attempt is made to propose empirical design equations to predict the ultimate strength of PVC-concrete composite columns (a new type of composite columns consisting of a PVC tube filled with concrete...

Versatile Aspects of IoT in Medical Science

With the advanced world, the technological benefits are increasing rapidly day by day. Internet of thing is an invention of modern world which is making our life easier and quicker in many aspects. Internet of thing has...

The Use of Automation Applications for Efficient Braking of Modern Wind Turbines

The purpose of this research is to study the wind turbine brake system in order to describe the structure, function and efficiency of automation that slows down and / or immobilizes a wind turbine in highwind conditions...

A Brief Review on Bio-Medical Application of Radar

Since, The Radar Was Invented It Was Associated With The Military And Defenceapplications. The Revolution Ofradar Then Offered Medical And Industrial Application Which Ledto Intelligent Environment To Monitor And React O...

Download PDF file
  • EP ID EP403319
  • DOI -
  • Views 89
  • Downloads 0

How To Cite

Pravin G. Sarpate, Ramesh R. Manza (2017). Speaker Recognition Using MFCC &Evolution with Different Classification Techniques. International Journal of Engineering and Science Invention, 6(9), 19-25. https://europub.co.uk/articles/-A-403319