Verifikasi Suara menggunakan Jaringan Syaraf Tiruan dan Ekstraksi Ciri Mel Frequency Cepstral Coefficient
Abstrak: Voice recording is an
important part of the evidence for the suspect, so it is necessary to verify
the voice suspects to prove the allegations of the suspect. The research aims
to develop a voice verification system using artificial neural networks and
extraction characteristics mel frequency cepstral coefficient. As the input
data analyzed is the data of the unrecognized voice recorder of the owner and
the recorded data of the sound that the owner has known as the comparison data.
Data input is processed by feature extraction consisting of framing, windowing,
fast Fourier transform, mel frequency wrapping, discrete cosine transform
resulting in mel-frequency wrapping coefficient. The mel frequency wrapping
coefficient of each frame in each input voice, is used as input on pattern
recognition using artificial neural networks. The results of artificial neural
networks are analyzed using decision logic to get a decision whether these two
voices are the same or not. The output of the system is a decision that the
tested sound is the same as or not with a voice comparison. Based on the level
of compatibility of the test data produces a voice verification system with
mel-frequency wrapping and artificial neural networks have a rate of 96%
accuracy. The accuracy of the voice verification system can be an option to
help resolve the issues in verification of voice recordings.
Keywords: Voice Verification
System; Mel Frequency Cepstral Coefficient; Mel-Frequency Wrapping; Artificial
Neural Network
Penulis: Andi Kurniawan
Kode Jurnal: jptinformatikadd170126