Ambisonics Soundfield Reproduction Based On Sparse Representation

As one of the three mainstream 3D audio techniques——Ambisonics system is respected by most of researchers for its recording system and reproduction system are independence of each other.The basis theory of Ambisonics system is that soundfield can be decomposed and reconstructed by a series of spherical harmonics,and the reproduction accuracy is positively associated with the order of harmonics function.However,in the recording process,the required number of microphones is very large when recording the higher order of soundfiled coefficients,especially in high frequency and large area.Based on the problem attention above,this paper introduced the thesis of compressed sensing,in order to reduce the number of required microphone.Furthermore,this method was applied to the soundfiled reproduction in reverberant room,especially in the measurement of the acoustic transfer function.The main work of this paper is embodied as follows:(1)In this paper,the method of soundfiled reproduction based on Ambisonics is studied in depth.Based on several soundfield reproduction methods including mode matching approach and continuous loudspeaker array hypothesis approach,the detailed derivation and general expression of the loudspeaker excitation signal are given,which all can reproduce the 3D soundfield effectively.In order to solve the problem that matrix inversion leads to unstable solutions due to the ill-conditioning of matrix when deriving the loudspeaker driving signal based on mode-matching method,which will reduce the quality of soundfiled reproduction.Based on the compressed sensing techniques,the sparse decomposition algorithm is introduced,in order to obtain stable solutions and reduce soundfield reproduction error.(2)In this paper,a method of 3D soundfiled recording based on sparse soundfield representation is proposed.The key of 3D soundfiled recording is how to obtain the harmonic components or the coefficients of soundflied.According to theoretical basis of Ambisonics,the coefficients can be calculated by the soundfield pressure which are measured by spherical microphone array.However,the frequency and area of sound signal that can be recorded accurately are limited by the number of microphones.Morever,the influence caused by Bessel zeros is analyzed in detail.Based on the general model of soundfiled,soundfield has sparse characteristics in the source domain,the sparse decomposition method is applied to obtain the sparse signal,and then upscale the harmonic coefficients to higher order,and then achieve the purpose of reducing the required number of microphones.(3)In this paper,a method of 3D soundfiled reconstruction in reverberant room based on sparse soundfield representation is proposed.Reverberation is an unavoidable problem in real scene.The ATF changes rapidly with the change of thesound source and the receiving point,the measurement of ATF in reverberation room is a rather tedious process.In addition,the key problem in the process of measurement is how to obtain the high order spherical harmonic coefficient of ATF with limited number of microphones.Meanwhile,the measurement equipment noise is also an important factor that affecting the accurate measurement of ATF.In this paper,based on the general model of ATF,and assume that the ATF is sparse in the source domain,the high order ATF coefficients are obtained by using the compressed sensing technique,which reduces the interference of the Bessel function zeros to the measurement stability.Experimental results show that,firstly,the proposed method for loudspeaker excitation signal optimization based on sparse representation solved the ill conditioned problem caused by matrix inversion,therefore,the sound field reproduction error of the listening area is reduced,and the effective range of sound is enlarged.Secondly,based on the sparse representation,the Ambisonics soundfield recording method in the free field greatly reduce the required number of microphones under the condition of ensure the accuracy of recording.Finally,the proposed method based on the sparse representation to measure the ATF,on the one hand,can reduce the number of microphones,on the other hand,can reduce the influence of the measurement device noise,meanwhile has a better measurement stability.