Blind score normalization method for PLDA based speaker recognition
This work addresses a specific problem in speaker recognition for applications like security or authentication, but it is incremental as it builds on existing PLDA methods.
The paper tackled performance degradation in PLDA-based speaker recognition when enrollment data sizes vary between speakers by introducing a blind score normalization technique that requires no extra development data. Experiments on the NIST SRE 2014 database showed improved accuracy in mixed enrollment conditions, with the method proven optimal in terms of detection cost function.
Probabilistic Linear Discriminant Analysis (PLDA) has become state-of-the-art method for modeling $i$-vector space in speaker recognition task. However the performance degradation is observed if enrollment data size differs from one speaker to another. This paper presents a solution to such problem by introducing new PLDA scoring normalization technique. Normalization parameters are derived in a blind way, so that, unlike traditional \textit{ZT-norm}, no extra development data is required. Moreover, proposed method has shown to be optimal in terms of detection cost function. The experiments conducted on NIST SRE 2014 database demonstrate an improved accuracy in a mixed enrollment number condition.