Proceedings of The 2nd International Conference On Advance And Scientific Innovation, ICASI 2019, 18 July, Banda Aceh, Indonesia

Research Article

Voice Identification Using Neural Network And Mel Frequency Cepstrum Coefficients

Download131 downloads
  • @INPROCEEDINGS{10.4108/eai.18-7-2019.2288519,
        author={Corianti  GMS and Fahmi  Fahmi and Maksum  Pinem and Sihar P. Panjaitan and Suherman  Suherman},
        title={Voice Identification Using Neural Network And Mel Frequency Cepstrum Coefficients},
        proceedings={Proceedings of The 2nd International Conference On Advance And Scientific Innovation, ICASI 2019, 18 July, Banda Aceh,  Indonesia},
        publisher={EAI},
        proceedings_a={ICASI},
        year={2019},
        month={11},
        keywords={voice recognition mel frequency cestrum coefficients artificial neural network},
        doi={10.4108/eai.18-7-2019.2288519}
    }
    
  • Corianti GMS
    Fahmi Fahmi
    Maksum Pinem
    Sihar P. Panjaitan
    Suherman Suherman
    Year: 2019
    Voice Identification Using Neural Network And Mel Frequency Cepstrum Coefficients
    ICASI
    EAI
    DOI: 10.4108/eai.18-7-2019.2288519
Corianti GMS1,*, Fahmi Fahmi1, Maksum Pinem1, Sihar P. Panjaitan1, Suherman Suherman1
  • 1: Electrical Engineering Department, Universitas Sumatera Utara, Indonesia
*Contact email: coriantisimamora@gmail.com

Abstract

Voice as a communication media for human and computer communication is developed by using voice recognition/identification technology. One of its applications is differentiate between male and female voices. Some speech recognition research have been proposed in disseminating male and female voices, this paper performs male and female voice extraction by using Mel Frequency Cestrum Coefficients (MFCC) as the characteristic vector in back propagation artificial neural network (ANN). The weight change cycle or EPOCH (Exponential Decay) is used to initialize male and female voice identification by using multiple recorded voices. As results, the female voice identification is better than of male with error less than 1.