Research On Spatial Cues Reproduction In Three-dimensional Sound System

Three-dimensional(3D)movies and virtual reality show realistic visual perception for users.However,immersive perception is not well reproduced due to the inconsistency of object's location shown by auditory and visual sense.The 3D audio standard proposal from MPEG inspired 3D sound research and development.3D audio system has three vital cues of a sound event:horizontal,vertical,and distance.The entire sound immersive perception depends on the above-mentioned three cues.With the rapid development of virtual reality,researchers and standardization organizations have invested a lot of effort and time into 3D sound systems.3D audio technology has become a hot topic in audio field.22.2 multichannel 3D sound system,designed by NHK,is a 3D audio standard for Ultra High Definition TV(UHDTV)in Japan.And it is a reference standard in MPEG-H 3D audio standard proposed by ISO/MPEG.Though researchers pay lots of attention to 3D audio,there are still several challenges:azimuth and elevation are well reproduced while distance is far from satisfactory.The computation complexity of 3D sound reproduction is rather high.And the prerequisite of reproduction method is rather strict.This paper focuses on distance cue reproduction in 3D audio systems,3D sound reproduction technique with no constrain,and efficient 3D sound reproduction method.(1)Particle velocity-based amplitude panning system(PVBAP)Amplitude Panning(PA)system is widely used by Dolby,DTS and MPEG-H 3D audio system.However,the classic amplitude panning methods require all the loudspeakers are on a spherical surface and listeners must locate at the center of loudspeaker array.This condition is hardly satisfied in the real reproduction environment.The direction of sound particle velocity(PV)is the same as the direction of sound and the amplitude of PV affects sound energy.This paper develops a particle velocity-based amplitude panning model to reproduce sound spatial perception.(2)Distance perception reproduction method in 3D sound system3D sound reproduction system needs a plenty of loudspeakers.This requirement is ill-equipped in ordinary domestic environment.Down-mixing method is needed to reduce the number of loudspeakers.This paper analyzes the sound energy error performance in the previous study.And then a energy maintaining model is proposed.We prove the minimum requirement about the numbers of loudspeakers and show analytic solution to reproduce distance and direction perception.Subjective and objective tests are conducted to show the performance of the proposed model.(3)3D sound reproduction method based on step-wise regression modelThe computational complexity of classical 3D sound reproduction theory such as Wave Field Synthesis(WFS),Ambisonics is rather high.In order to efficiently reproduce 3D sound field,this paper analyzes the similarity between sound field reproduction and machine learning method and proposes a step-wise regression model to reproduce 3D sound field.The model performance parameters such as MSE,RMSE,SSE,raw residual,Pearson residual,standard residual are tiny.The proposed model is flexible for theatre and personal environments to produce 3D sound for users.