Research On Binaural Ambisonics Reproduction System

Ambisonics is a series of spatial sound reproduction system based on spatial harmonics decomposition and each order approximation of sound field.Binaural Ambisonics is a binaural rendering technique that combines the principles of physical sound field reconstruction and binaural sound pressure reconstruction.By using head-related transfer functions(HRTFs)filters,binaural Ambisonics converts the Ambisonics signals for headphone reproduction to create auditory perceptions similar to those of Ambisonics reproduction.It can be used not only to convert the Ambisonics signals for headphone reproduction,but also to directly generate binaural signals for headphone reproduction.It is one of hot topics of spatial sound reproduction in recent years,and has been included in the standard of MPEG-H 3D Audio by ISO and IEC.Although there have been some researches on this topic,binaural Ambisonics is still a technique being developing and some problems are needed to be addressed.First,most of the current researches are focused on theoretical analysis or computational simulation,experiments on auditory performances are usually limited to the cases of static binaural Ambisoncis reproduction,and there are only several reaearch groups who really realize dynamic binaural Ambisonics reproduction.Second,there is still lack of a design criterion for binaural Ambisonics.In addition,restricted by the Shannon–Nyquist spatial sampling theorem,a very high order binaural Ambisonics is needed to synthesize the correct high frequency spectral cue at binaural pressures.This causes difficulty in practical use.On the other hand,dynamic cue caused by head turning may replace the high frequency spectral cue to provide information for vertical localization perceptions.However,the effect of the dynamic cue on the vertical localization is not fully clear.To address the above problems,following works have been carried out in this thesis:Firstly,a series of virtual source localization experiments was conducted to examine the effects of dynamic and spectral cues on auditory vertical localization.The results indicated that both dynamic and spectral cues contribute to vertical localization.However,the information provided by the two cues is somewhat redundant.When one cue is eliminated,another cue alone still enables vertical localization to some extent.When both cues are eliminated,vertical localization is impossible.A coordination of the dynamic and spectral cue,especially the individualized spectral cue,enhances the vertical auditory localization.These conclusions impact on the design of binaural Ambisonics.Secondly,a PC-based real-time binaural Ambisonics experimental platform was set up.The dynamic performance of this platform satisfies the requirements of scientific experiments,and the parameters of binaural Ambisonics,such as the number of virtual loudspeakers and the order,can be chosen flexibly.Based on this platform,evaluation of dynamic binaural Ambisoncis and other related works can be conducted.Therefore,this platform serves as a reference system for the application of binaural Ambisonics.Then,the effects of the number of virtual loudspeakers and the order on sound pressure error for the binaural Ambisonics reproduction were analyzed,and subsequently verified by psychoacoustic experiments.Based on the analytical and experimental results,the design criterion for dynamic binaural Ambisonics reproduction was proposed.It is suggested that a dynamic binaural Ambisonics with order(L-1)=3~5 and the number of virtual loudspeakes being slightly higher than(L+1)~2 is enough to create appropriate directional localization performances.Of course,in order to further improve the localization perception,a higher order can be adopted,but the reproduction system will be more complex.Fourthly,a method of binaural near-field compensation high-order Ambisonics was proposed.This method combines the principles of virtual auditory display and the near-field compensation high-order Ambisonics,so as to to create or control the distance and direction perceptions of near-field virtual source in headphone reproduction.And the psychoacoustic experiment verified that binaural near-field compensation high-order Ambisonics with order(L-1)=5 is able to reate appropriate distance and direction localization effects.Finally,spherical microphone array recording and binaural rendering(SMARBR)is implemented.The effects of the number of virtual loudspeakers and the order on sound pressure error for SMARBR were also analyzed,and verified by psychoacoustic experiments.The results show that although there exists extra error caused by sound field recording in SMARBR,the effects of the number of virtual loudspeakers and the order are similar to that of the directly synthetic binaural Ambisonics signals.This work is a relatively systematic and comprehensive study of binaural Ambisonics.It is also an important reference for the applications of binaural Ambisonics.