The Research On Speech Signal Dereverberation Technology

In daily life,when using telephones,hearing aids,mobile phones and other equipments in some relatively enclosed indoor space,an intense reverberation phenomenon will produce when the location of the sound source is far away from the sound receiver.The existence of reverberation results a masking effect between the two or more syllables,seriously affecting the auditory effect.Speech signal dereverberation processing as an important part of the speech enhancement.Simultaneously,it provides early preprocessing with speech signal processing such as speech synthesis,sound source localization,speech recognition.In addition,speech signal dereverberation related technology is also widely used in other aspects such as architecture acoustics,vibration acoustics,seismic data analysis,biomedicine,radar and sonar.The following is the main contents of the research work of the thesis:Firstly,the theories of reverberation are analyzed in this thesis,such as the generation of reverberation,mathematical model and characteristic parameters,and studies the evaluation indicators of speech dereverberation performance,including two subjective evaluation indicators and three objective evaluation indicators.At the same time,a large number of simulation experiments prove that two objective evaluation methods in frequency domain are more applicable to reflect subjective sense than the objective evaluation method in time domain.Secondly,the related theories of the complex cepstrum are analyzed in this thesis.Under the condition of single channel reverberation,respectively studies the complex cepstrum domain filtering speech dereverberation method,speech dereverberation method based on minimum phase decomposition and complex cepstrum blind deconvolution speech dereverberation method,and carries out the simulation analysis of this three methods.Finally,for single microphone speech dereverberation technology only use time domain and frequency domain information,but available space information is limited,it is hard to gain a better dereverberation effect,this thesis studies the microphone array speech dereverberation technology,and combines beamforming technologies with other single channel speech dereverberation technologies.First,the DSB speech dereverberation method based on fixed beamforming and the TF-GSC speech dereverberation method based on adaptive beamforming are studied,and carries out the simulation analysis of the DSB and TF-GSC method,for their good inhibitory effect on additive noise and beam direction dereverberation,combines DSB with complex cepstrum blind deconvolution technology,obtains a more effective dereverberation method.Combines TF-GSC with minimum phase decomposition technology,and for speech dereverberation method based on TF-GSC and minimum phase decomposition have the problem of large amount of calculation,analyzes the improved speech dereverberation method based on TF-GSC and minimum phase decomposition.this method makes use of the phase of reverberation speech signal received by a microphone replace the phase information of the all-pass component in order to reduce the amount of calculation.And through comparing the simulation results,combines the beamforming technologies with other single channel speech dereverberation methods obtain good dereverberation effect.