Dissertation > Excellent graduate degree dissertation topics show

Research and Implementation of Voice Noise Reduction Technology

Author: DengLiNa
Tutor: HuangXiaoGe
School: University of Electronic Science and Technology
Course: Access to information and detection technology
Keywords: Voice Noise Reduction Spectral subtraction Genetic Algorithms Digital trap
CLC: TN912.3
Type: Master's thesis
Year: 2011
Downloads: 152
Quote: 1
Read: Download Dissertation

Abstract


In a real life environment, voice signals will be in the process of encoding, transmission, etc. The various types of noise pollution, noise has been the main reason for the deterioration of the voice processing system performance dramatically. Speech noise reduction is an effective treatment technology for the noise problem, and its purpose is to eliminate the impact of noise and improve speech intelligibility, improve the quality of the voice. Currently, voice noise reduction algorithm. The spectrum subtraction principle is simple, easy to achieve the advantages of common algorithms for speech and noise reduction. Spectral subtraction has two shortcomings: First, the spectrum subtraction performance is good or bad depends on the noise estimation, noise estimation depends on the endpoint detection algorithm. Usually the strength of the noise level is high, the endpoint detection algorithm will fail, the specific location can not be detected by the signal noise frame, thereby affecting the accuracy of the noise estimate value; second signal with noise after a spectral subtraction noise reduction, since the subtracted spectral subtraction is the same noise estimated value, so that the signal will appear randomly separated spectral region, these spectral region formed a \For the above two shortcomings of the spectrum subtraction, its improvement. First: in order to make the noise endpoint detection algorithm can correct detection when the noise level is high, we ask with the mean amplitude value of the noise signal with noise, and to judge according to the mean of the mean noisy signal the beginning of the magnitude of the number of frame size the beginning of the signal is noise or noisy voice signal. And then according to the change of the difference of two consecutive frames of the signal to determine the starting position of the noise frames and speech frames, the same time while judgment to obtain the mean value as a noise estimate value, so that both take into account the continuous before and after the two signals related is capable of attenuating noise. In addition, based on the noise estimate value of the noise endpoint detection method improved in this paper can be on the entire band noisy speech signal a fast update the noise estimate value, to improve the spectral subtraction real-time processing capability. Second: In order to reduce the the musical noise introduced by spectral subtraction, we realize the LMS algorithm in the time domain speech enhancement processing spectral subtraction noise reduction signal. LMS algorithm to reduce the noise level at the same time, musical noise is converted to white noise is of lower energy, reduce musical noise stimulus to the human ear, to help improve the voice quality of the processed audio improve the objective and subjective evaluation results. In addition to improved spectral subtraction for containing solid-frequency noise in the noisy speech signal design combination of genetic algorithms and digital voice notch filter and noise reduction algorithms. As long as we Noisy Speech select a period containing solid-frequency noise signal as a reference signal, the reference signal by genetic algorithm search to obtain accurate frequency value of solid-frequency noise. On this basis, we designed the IIR digital notch filter. IIR digital notch filter will not only be able to filter the specific frequency signal, and little attenuation of the signal of other frequencies, it is possible to protect the voice signal. IIR digital notch filter with fixed frequency noise component in the noise signal in the filter processing is performed to achieve the purpose of noise reduction of speech.

Related Dissertations

  1. Voice Signal De-noise Base on Wavelet and Arithmetic Implement with DSP,TN912.3
  2. Development of the on-line Training and Examination System of Army,TP311.52
  3. Designs and Applications of Fuzzy Synthetic Evaluation Models Based on Parallel Algorithms,TP18
  4. Genetic Algorithm in logistics and warehousing Optimization Research,F259.2
  5. Mining resources based on genetic algorithm optimization model of,O224
  6. Band Entropy Method and Its Application to Fault Diagnosis of Rolling Bearings,TH165.3
  7. Based on Human Auditory Masking spectral subtraction algorithm and implementation,TN912.3
  8. 120 Tons Gantry Crane Structural Analysis and Girder Structural Optimization Based on Genetic Algorithm,TH213.5
  9. Research of Digital Image Segmentation,TP391.41
  10. The Research and Design of the Intelligent Electronic Belt Scales,TH715.195
  11. Based on the angle of the knee walker functional electrical stimulation Fuzzy Control,R651.2
  12. The Topology Optimal Study of Truss Structures Based on Improved Genetic Algorithm,TP18
  13. Based on hierarchical task network planning optimization methods for military,E072
  14. Evolutionary Logic Research in Artificial Intelligence,B812.3
  15. Study on Back Analysis Method with Genetic Algorithm of the Deep Foundation Pit’s Design Parameters and Its Excavation Dynamic Prediction,TU473.2
  16. The Research of Optimize Numerical Blending Based on Artificial Neural Networks,TS41
  17. Application and Performance of WFGD System of Pannan Power Plant,X701.3
  18. Application of Wireless Sensor Nets in Environment Monitor Svstem,TN929.5;TP212.9
  19. The Optimal Portfolio Decision Model with Mean-VaR Risk Control,F830.9
  20. Fingerprint recognition algorithm research related,TP391.41
  21. Robot obstacle avoidance algorithm,TP242

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