Study On Sound Synthesis Of Liquid-solid Interaction Scenes

In recent years,sound synthesis technology for virtual scene has developed rapidly,which has received widespread attention in the field of computer graphics and virtual reality,and has been widely used in video game,film,studio entertainment and other fields.However,most of the current sound synthesis methods focus on the sound model construction of single form sound source such as solid,liquid and gas.The sound synthesis technology of highly non-linear sound source such as liquid-solid interaction has not been carried out.In this thesis,we study the construction method for the liquid-solid interaction scene,and explore the key technical problems in the process of synthesizing liquid-solid interaction sound,including difficult detail preservation,high computational complexity and difficult collaborative interaction.The detailed contents are as follow:(1)Detail-preserved liquid-solid interaction sound synthesis based on sound texture transferAiming at the key technical problem of detail preservation,we propose a detail preserving sound model for liquid-solid interaction scenes which combines physical acoustic model and data-driven method.In order to solve the problem of missing sound details caused by the mismatch of audio-visual parameters in the process of sound modelling,a spectrogram texture transfer algorithm is designed.Moreover,a fine-grained coupling synchronization algorithm based on key frame fusion is designed to solve the problem of synchronization difficulty caused by highly non-linear dynamic coupling.The proposed sound model can keep timbre details and audio-visual synchronization details,and provides a sound synthesis solution for the liquid-solid interaction scenes with highly discriminable sound details and complex motion.(2)Efficient computational liquid-solid interaction sound synthesis based on statistical-based physical modelAiming at the key technical problem of high computational complexity,an efficient calculation sound model is proposed by combining the physical model and statistical model.A material shader algorithm based on signal decomposition and reconstruction is proposed to solve the problem of time-consuming sound material modelling.To settle the co-ordination problem of multi-source with randomness distribution,a dynamic activation algorithm is proposed to balance the computational complexity and simulation authenticity.By combining the statistical method,the proposed sound model can avoid the problem of repeated calculation and realize the high-efficiency synthesis of liquidsolid interaction sound in large,multi-source scene.(3)Collaborative interactive liquid-solid interaction sound synthesis based on force-driven modelAiming at the key technical problem of cooperative interaction,a liquid-solid interaction cooperative interactive model is proposed which integrates haptic information and visual information.It is difficult to find a real-time solution since the two-way coupling increases the complexity of liquid-solid interaction.To solve this,we design a spectral granular sound synthesis algorithm for real-time feedback.In order to solve the problem of difficult mapping construction caused by the large difference of sampling rate of different perception channels,based on the idea that haptic feedback and auditory feedback follow the same physical principle,a homology mapping algorithm is proposed which realizes the accurate synchronization of synthetic sound and other perception channels.In conclusion,we focus on the liquid-solid interaction sound synthesis method and explore the three aspects of detail preservation,efficient calculation and collaborative interaction,which solves key technical problems in the liquid-solid interaction sound synthesis.This research provides algorithmic foundation for constructing a highly realistic and immersive virtual reality system,which can effectively improve the human-computer interaction effect for military exercises,education,film production,game production,virtual reality and other fields.