Klasifikasi Usia Berdasarkan Suara Dengan Ekstraksi Ciri Mel Frequency Cepstral Coefficients Menggunakan Support Vector Machine

Mufi Oktaviani (Universitas Islan Negeri Syarif Hidayatullah Jakarta, Indonesia)
Taufik Edy Sutanto (Universitas Islan Negeri Syarif Hidayatullah Jakarta, Indonesia)
M Mahmudi (Universitas Islan Negeri Syarif Hidayatullah Jakarta, Indonesia)

Kesatria : Jurnal Penerapan Sistem Informasi (Komputer dan Manajemen)
Vol 4, No 4 (2023): Edisi Oktober

Each individual has a different and unique voice, this can potentially be used to determine a person's age. Classification can be done through a feature extraction process using Mel Frequency Cepstral Coefficients (MFCC). Then the extraction results can be used as features in machine learning models such as Support Vector Machines. Data was collected by recording 100 voices of informants, then the feature extraction process was carried out using MFCC, and then classified using Support Vector Machine. Age categories set are children (0-14), teenagers (15-40) and elderly (40+). Then the data samples were duplicated for variety using noise, shift and dynamic change methods so that 6000 sound samples were obtained. Our model can result in an accuracy of 75,8% and the recall obtained in the children category is a percentage of 80%. Researchers hope that these results can be a reference for human age recognition application systems for voice-based classification.