Within-Class Covariance Normalization (WCCN) for Channel Normalization Based of Indonesian Speaker Recognition System

Speaker recognition system is a technology that can be used to verify the speaker's identity from an unknown speech sample. In Indonesia, this system is actively used to assist the speaker verification process as an evidence in court. The speaker recognition system developed in this study is based on i-vector modeling, along with the usage of cosine distance scoring. This system is trained and tested using a speech database spoken in Indonesian language. Previous study using same database result in 3.50% equal error rate (EER). This study perform an effort to improve system performance in the presence of mismatch channels using within-class covariance normalization (WCCN) technique. Through this, the system has successfully lower its EER value to 2.40%.