Dissertation > Excellent graduate degree dissertation topics show

Multiple ANN/HMM Hybrid Used in Speech Recognition

Author: LiuMingYu
Tutor: LiHaiFeng
School: Harbin Institute of Technology
Course: Computer Science and Technology
Keywords: speech recognition ANN/HMM optimize of state number multiple hybrid model method of reorganizing features adaptively
CLC: TN912.34
Type: Master's thesis
Year: 2008
Downloads: 166
Quote: 0
Read: Download Dissertation

Abstract


Speech is the most natural and familiar interactive way for human,and current days, speech recognition and speech synthesis are ascendant.In the case of adequate training samples, isolated word recognition has got very gratifying achievements. In some cases, however, the number of training samples may not see the need for training model. In order to obtain acceptable recognition rate the model must be improved further. Based on the original timing model which combines Artificial Neural Networks and Hidden Markov Model(ANN/HMM), this paper studies a multiple and hybrid recognition method to get high rate by establishing model complementary for defferent features.Artificial Neural Network (ANN) is used as model int the status class with features of anti-noise, anti-variant, adaptive, learning ability, high recognition speed,and is also the model of the basic unit of the object to be recognised. As the model of the whole pattern, Hidden Markov Model (HMM) has strong ability to deal with time-series. In this method, the combination of ANN and HMM is on the frame level. The output error of ANN is used to estimate output probability of one state of HMM. Furthermore, a method of auto-split-and-merge the state number is used to determine the state number of a model.In this method,states are automatically added or deleted on a proper position according to the training data.We split the states with low modeling precision,delete the redundant ones,and finally achieve a balance.On the basis of above model, we propose a multiple ANN/HMM hybrid model, which segments features with competitive learning mechanism and reduces the cost of storage and calculation of systems with the method of reorganizing features adaptively.This method can use the adaptive learning capacity of ANN to ensure the system’s good performance.Taking speech commands for example,we compare the modeling effects between this method and traditional ones.The results show that this multiple model can improve the modeling precision and rate not with consuming resources of system massively.In order to put the research achievement into use,we developed a simple multi-mode Human-Machine Interaction system.In this system,we can speak to give orders to the computer in a more natural way. With the use of this system, it has the characteristics of a fast response and high recognition rate.

Related Dissertations

  1. The Design of a DSP-Based Robot Speech Command Recgnition System,TN912.34
  2. The Design and Research of Health Management Based on Smartphone Environment,TN929.53
  3. Research on Hmm-based Speech Recognition System of the Robot,TN912.34
  4. MFCC -based speech recognition system to improve research and design,TN912.34
  5. Research and Implemenation of Voice Intelligent plantform Based on VoiceXML,TP311.52
  6. Topic Classification of Speech Documents Based on the Word Fragment Network,TN912.3
  7. Study on Hybrid Model of Speech Recognition Based on HMM and PNN,TN912.34
  8. Mobile robot voice recognition control simulation system design and implementation,TN912.34
  9. Research on DBN-Based Continuous Speech Recognition,TN912.34
  10. STRAIGHT spectrum - based speech recognition algorithm research,TN912.34
  11. Research on the Key Technologies fo Speech Recognition for Robot Communication,TN912.34
  12. The LVCSR system based on adaptive methods of semi-supervised learning,TN912.34
  13. Parallel Optimization Method in Language Model for Mandarin Speech Recognition,TN912.34
  14. Research of Segmentation Based Chinese Continuous Speech Recognition Technology,TN912.34
  15. National language language recognition research based on support vector machine,TN912.34
  16. Phone-level Based Mispronunciation Automatic Detection and Its Application,TN912.34
  17. A Design of Intelligent Terminal of Voice Control Based on ARM9,TN912.3
  18. The Research and Implementation of Algorithm of Isolated Word Speech Recognition,TN912.34
  19. Distributed Speech Recognition and Voice XML Standardlanguage in Vivid-Ring Application,TN912.34
  20. Application of Speech Recognition in the Testing System of Electromagnetic Valve,TN912.34

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