Text Independent Speaker Identification for Myanmar Speech
Win Lai Lai PhyuWin Pa Pa
Natural Language Processing Lab., University of Computer Studies
摘要：Nowadays, speech signal processing is one of the emerging application areas of digital processing. There are many research areas related to speech processing such as speaker recognition, speech recognition, and speech synthesis. Speaker identification is the task of analyzing the speakers’ characteristics in speech to exactly identify individuals. The identification task performs better when there is enough background training data. Mel Frequency Cepstral Coefficient（MFCC）, Perceptual Linear Prediction（PLP） and Filter Bank features are extracted as front-end processing. Constructing Universal Background Model（UBM） is the main component of i-vector system as it is essential for collecting statistics from speech utterances and for clustering the speaker models. This paper indicates that the impacts of unlimited speech data in speaker identification by using i-vector method with probabilistic linear discriminative analysis（PLDA） approach and the import role of speaker models in identification process.
2019 The 11th International Conference on Future Computer and Communication （ICFCC 2019）