Educational project on a simple voice identification using mel-frequency cepstrum coefficients and vector quantization

• Main Authors: Akmeliawati, Rini, Widhiputranto, H. K.
• Format: Conference or Workshop Item
• Language: English
• Published: 2008
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://irep.iium.edu.my/14165/1/paper794.pdf; http://irep.iium.edu.my/14165/
• Institution: Universiti Islam Antarabangsa Malaysia

Voice identification is a task of identifying a person based on the content of information in one’s voice signal which distinguishes his/hers from voices of other people. This paper presents an educational project on a simple voice identification system, which uses two parameters of voice features: the fundamental frequency or pitch and the vocal tract information represented by Mel-Frequency Cepstrum Coefficients (MFCC). The estimation fundamental frequency is performed with Harmonic Product Spectrum (HPS) algorithm, whilst MFCC is calculated using Discrete Cosine Transform, then, is modeled with Vector Quantization to compress the size of data to be processed and to assist identification.