Dissertation > Excellent graduate degree dissertation topics show

Research on Text-independence of Open-set Speaker Recognition

Author: LuChunMei
Tutor: WangJianYing
School: Southwest Jiaotong University
Course: Signal and Information Processing
Keywords: Speaker Recognition Endpoint Detection Principal component analysis MFCC Open set identification Threshold confirmation
CLC: TN912.34
Type: Master's thesis
Year: 2011
Downloads: 43
Quote: 0
Read: Download Dissertation

Abstract


With the development of information technology, identification in the field of information security is playing an increasingly important role. Biometric its unique stability, uniqueness and convenience, has become an important field of identity research. Speaker recognition technology is a biometric technology, this technology is based on the focus of the speaker differences between personality traits to distinguish the speaker. Currently closed set text-related speech recognition technology has achieved a high recognition rate, but for the open set recognition, the recognition rate is low. Because identified to be involved in an open set speaker identification and recognition thresholds, so the identification and recognition is an open set threshold identification difficult and critical. In this paper, text-independent speaker recognition technology open sets, a detailed analysis of the speech recognition system, the basic principles and structure, and for speech endpoint detection, feature extraction, open set identification, recognition threshold depth of several parts research. The main work is as follows: (1) pre-processing and voice endpoint detection portion, the first pre-processing and analysis of the importance of speech endpoint detection. Then theoretically description is based on short-term energy endpoint detection algorithm, based on short-term zero rate endpoint detection algorithm, based on the value of short-term energy frequency endpoint detection algorithm based on spectral entropy endpoint detection algorithm and improved spectrum entropy endpoint detection algorithm. Finally, pre-emphasis and the five experimental endpoint detection algorithm simulation, and endpoint detection algorithm for the five comparative advantages and disadvantages. Recorded under laboratory conditions for speaker pronunciation smaller features, an improved spectral entropy endpoint detection algorithm. (2) feature extraction part, from the theoretical analysis of the parameter extraction method and PCA theory. PCA theory is applied to study the characteristic parameter extraction. Simulation results show that the method can improve the recognition performance to some extent, while reducing the computation time codebook training. (3) identify some open set, the lack of traditional VQ study the FCM algorithm combined with PCA theory open set speaker identification system. Simulation results show that FCM PCA than the FCM, VQ PCA and VQ higher recognition rate. (4) open set identification section, a detailed description of the classic threshold, dynamic threshold, RS threshold estimation method. FCM PCA study will identify and confirm RS threshold combining open set speaker recognition system. Simulation results show that this system is based on the three EER threshold EER recognition system compared to a certain extent reduced.

Related Dissertations

  1. Application of Improved Principal Component Analysis Algorithm in Course Construction,G642.4
  2. Research of Diagnosing Cucumber Diseases Based on Hyperspectral Imaging,S436.421
  3. The Impact of Tourism on Typical Vegetation in Luya Mountain Nature Reserve, Shanxi Province,S759.9
  4. Macaca mulatta palm morphological study of pattern ridge count,Q954
  5. Zhaoguan Lower Coal Group water inrush prediction and control techniques,TD745
  6. Research on Cultural Industrial Competitiveness of Chong Qing,F224
  7. The Research of Prairie Road Light Environment Effects on Physiological Indicators of Drivers,U491.254
  8. Research on Feature Extraction, Selection and Classification Algorithms for Pulmonary CAD,TP391.41
  9. The Research and Empirical Analysis on the Relation between Financial Ecology and Economic Growth in Jiangsu Province,F127;F224
  10. Research Onamethod for Human Face Recognition Based on MMTD,TP391.41
  11. Competitiveness’ Evaluation and Development Strategy of the Tourism Industry in Huanggang City,F592.7
  12. Face recognition algorithm based on feature fusion research,TP391.41
  13. Image semantic annotation of blocks - a global feature extraction method,TP391.41
  14. Waste Road density points seismic signal automatic identification and picking first method,P631.4
  15. Power spectrum estimation in the broadband ADCP Signal Detection Research and Application,TN911.23
  16. Integrated Intelligent Computing Based Fault Diagnosis of Analog Circuits,TN710
  17. The Research on Microscopic Evaluation System of Road Traffic Safety,U491
  18. Research on Speaker Recognition System under the Visual C++ 6.0,TN912.34
  19. Ability of local governments to finance the construction of evaluation system,F812.7
  20. Research of Commodity Index Which Based on Dynamic Adaptive Hedonic Model,F713.35
  21. Effect of Roasting on Camellia Oil Aqueous Extraction Technique and Its Quality and Aroma,TS224

CLC: > Industrial Technology > Radio electronics, telecommunications technology > Communicate > Electro-acoustic technology and speech signal processing > Speech Signal Processing > Speech Recognition and equipment
© 2012 www.DissertationTopic.Net  Mobile