Environment based threshold for Speaker Identification

Abstract

Speaker Identification process is to identify a particular vocal cord from a set of existing speakers. In the speaker identification processes, the unknown speaker voice sample targets each of the existing speakers in the system and gives a predication. The predication is more than one existing known speaker voice and is very close to the unknown speaker voice. It is a one to many mapping. The mapping function gives a set of predicated values associated with the order pair of speakers. In the order pair, the first coordinate is the unknown speaker and the second coordinates is the existing known speaker from the speaker recognition system. The set of predicated values helps to identify the unknown speaker. The identification process makes a comparison of the unknown speaker model with the models of the existing voice in the system. In this paper, the model is a Gaussian mixture model built by the extraction of the acoustic feature vectors. This paper presents the impact of the decision threshold based on false accepts and false reject for an unknown number of speaker conversion in the speaker identification result. In the simulation, the considered known speaker voices are collected through different channels. In the testing, the GMM voice models of the known speakers are distributed among the numbers of clusters in the test data set.