Dissertation > Excellent graduate degree dissertation topics show

The Study of Speech Endpoint Detection Methods Based on Cepstrum Characteristic and Voicing Features

Author: SunHaiYing
Tutor: LiuYun
School: Qingdao University of Science and Technology
Course: Control Theory and Control Engineering
Keywords: speech signal endpoint detection cepstrum characteristic voicing feature
CLC: TN912.3
Type: Master's thesis
Year: 2008
Downloads: 282
Quote: 1
Read: Download Dissertation

Abstract


Endpoint detection serves as the core technology in the preprocessor of the speech recognition system, its directly determines the succeed of the whole speech recognition system to a great extent. As we know, the object of speech signal processing is effective speech signal in facts, so it is essential for us to find the begin and end point of the effective speech signal in input signal. Endpoint detection aims at finding the begin and end point of the speech signal in input signal, we could also call that dynamic speech detection.In this paper, it first presented bring-model of the speech signal in briefly, including the characteristic of speech signal and noise signal, some theory about buffer window, separate frame processing of speech signal and so on. Second, it introduced some endpoint detection methods proposed by the domestic and international scholars in recent years, and analysed about detection methods with simulation results given in particular. Afterwards, it introduced a estimate method which can manifest periods of the speech voicing features. Finally, the ameliorative speech endpoint detection algorithm adopted in this article is proposed, including:One, endpoint detection base on cepstrum characteristic (cepstrum distance) is executed with the speech signal given, and save the information into cache memory.Two, the methods of detecting the about the speech signal are adopted, including detecting the speech harmonic information of the speech signal base on speech matrix chart, the voicing features base on the speech periods estimated and the method of detection voice and unvoice base on the wavelet transform.After analysis and simulation experiment, we can find that the endpoint detection method proposed in this paper can preferably detect the endpoint of speech signal, and have better veracity, outstanding noiseproof feature, strong robustness.

Related Dissertations

  1. Power spectrum estimation in the broadband ADCP Signal Detection Research and Application,TN911.23
  2. Speech Emotion Recogonition Research,TN912.34
  3. Research and Implementation of Speech Recognition System for Mobile Robot,TN912.34
  4. Parameter Estimation for Non-stationary Speech Feature,TN912.3
  5. The Study of Speaker Recognition,TN912.34
  6. Application of Wavelet Packet Analysis on Speech De-Noising,TN912.3
  7. VQ with HMM model experimental study of the speech signal,TN912.3
  8. Research on Emotion Recognition of Speech Signal Based on HMM,TN912.34
  9. Research and Application of Pronunciation Detection for Deaf Children Rehabilitation,TN912.3
  10. Study on Voice Signal Filtering Based on DSP,TN912.3
  11. Speech Signal Processing Based on Auditory Neural Mechanisms,TN912.3
  12. Computer Analysis-based Pronunciation Quality Assessment in Language Learning System,TN912.3
  13. Research on Speech Signal Processing Technology,TN912.3
  14. Speech recognition technology in language teaching software,TN912.3
  15. Research of the Speech Recognition Technology Based on DSP,TN912.34
  16. Research and Implementation of Adaptive Low-Rate Speech Coding,TN912.3
  17. Endpoint Detection of Speech Signal Based on Empirical Mode Decomposition,TN912.3
  18. Telephone-based channel voiceprint recognition algorithm,TN912.34
  19. Research on Noisy Voice Activity Detection Method,TN912.3
  20. Noisy Speech Recognition System,TN912.34

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