Applications Of Time Frequency Analysis In Speech Signal Enhancement

Noise interference is ubiquitous in our lives,and it affects people's communication at every moment.Therefore,speech enhancement technology has always been a research hotspot.The speech signal that we generally know is a non-stationary signal,and time-frequency analysis is a powerful tool for studying non-stationary signals.Unlike traditional Gaussian noise,a ubiquitous interference cannot be described by a Gaussian distribution and is a more perturbative interference.Under the influence of impulsive interference,the performance of traditional speech denoising algorithms is not ideal.Therefore,we need to further study the problem of speech enhancement under impulsive interference.The main research contents are as follows:1.Research on the problem of de-interference for speech signals affected by impulse interference.The impulsive interference,in this work,is modeled by an unknown sparse vector so that it can be actively suppressed.The speech signal is sparsely represented by the wavelet domain.To achieve the simultaneous speech recovery and the noise suppression,a joint estimation is devised based on the fact they have sparse representations in different domains.To efficiently solve the problem,the joint greedy estimation algorithm and the alternating direction multiplier(ADMM)method are used to obtain solutions respectively.Simulation results demonstrate the superior performance of the proposed approach.2.The speech denoising problem in the presence of mixed impulsive interference and Gaussian noises is investigated by exploiting transform domains.To that end,the proposed noise suppression scheme is a cascaded form consisting of an impulsive noise suppression module and a Gaussian noise suppression module.For the impulsive interference reduction subsystem,in this work,the noise is sparsely represented by the time domain whereas short time Fourier transform(STFT),wavelet transform(WT)and wavelet synchrosqueezed transform(WSST)are studied to provide sparse representations for the speech.By utilizing the transform domains,the speech recovery and the impulsive interference suppression are simultaneously achieved under an optimization framework.Subsequently,the alternating direction method of multipliers(ADMM)is used to solve ?1-norm constrained optimization.In the Gaussian noise reduction subsystem,the Gaussian noise is suppressed by the famous Wiener filter in the transform domains as well.Numerical studies including simulations and real data analysis demonstrate the superior performance of the proposed scheme.Finally,the article analyzes the speech signal from the perspective of group sparse and tries a new method to deal with Gaussian noise.The simulation experiment also verifies the feasibility of the method.