Loading...

Text-Independent Speaker Identification in Large Population Applications

Zeinali, Hossein | 2012

589 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43393 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Sameti, Hossein
  7. Abstract:
  8. The human speech conveys much information such as semantic contents, emotion and even speaker identity. Our goal in this thesis is the task of text-independent speaker identification (SI) in large population applications. Identification (test) time has become one of the most important issues in recent real time systems. Identification time depends on the cost of likelihood computation between test features and registered speaker models. For real time application of SI, system must identify an unknown speaker quickly. Hence the conventional SI methods cannot be used. The main goal in this thesis is to propose several methods that reduced identification time without any loss of identification accuracy. First of all, the overall structure of the SI systems is introduced; followed by details of different parts of it. The Gaussian Mixture Model is explained as basis method for SI systems. Then, the common methods in identification time reduction is presented. After that, a two-step method that uses Sparse Representation in the first step is proposed. Our experiments on the TIMIT database show that by using this method, we can achieve 18× speed-ups without any loss of accuracy. In addition, another two-step method is proposed that use Models Distance Method in the first step. Experiments on TIMIT database show that this method can identify speakers 28 times faster than conventional method without any loss of accuracy. At the end, another two-step method that uses Nearest Neighbor distance is proposed. This method achieves 3.4× speed-ups without any loss of accuracy.
  9. Keywords:
  10. Text-Independent ; Sparse Representation ; Speaker Identification ; Real Time System ; Gaussian Mixture Modeling

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC