Research On Speech Enhancement Method Based On Phase Space Reconstruction

Voice is the most direct way for people to communicate in life.No matter how technology develops,voice is the most direct way for people to communicate with each other.However,the speech signal is interfered by noise.Due to the presence of noise,errors may occur during the transmission of information and affect the quality of voice communication between people.Therefore,the research on speech enhancement has gradually attracted people's attention.The primary task of this research is to remove the noise in the speech signal,and improve the speech quality.There are many classical speech enhancement techniques.This paper proposes a speech enhancement algorithm based on phase space reconstruction.The main research contents are as follows:1.This paper introduces the basis of speech enhancement,research significance,research status and so on.The characteristics of speech and noise are analyzed.And thejingdian classical speech enhancement algorithm is described.In order to compare the effects of different algorithms,two evaluation methods are described.2.The chaotic characteristics of speech signals are analyzed based on the phase space reconstruction theory.Firstly,the chaotic theory is introduced in detail,and the fractal dimension and Lyapunov exponent are calculated to prove the chaotic characteristics of the speech signal.On this basis,the appropriate parameters of correlation dimension and delay time are selected,and the phase space is reconstructed using Takens theorem.3.This paper is based on phase space reconstruction of speech signals.The appropriate tangent space for the reconstructed points to project is fund.First,the algorithm groups the points after phase space reconstruction and estimates the tangent space of each neighborhood.Then projection denoising is carried out to tangent space,and finally the denoised data is replaced by iterative processing.In this paper,multiple noises of noisex-92 library are added to serval signals of NOIZEUS library,and used in speech enhancement experiments.The objective evaluation method is used to calculate the signal-to-noise ratio of five algorithms.Experiments show that the tangent projection method based on phase space reconstruction improves the signal-to-noise ratio of speech signals by 5-15 db,and the filtering effect is better than other traditional algorithms.