A new algorithm enables more realistic sound effects in VR

Researchers at Stanford University have managed to create realistic sound effects for virtual worlds by dividing computing power by one thousand. An advance that will allow much more immersive sound environments.

The virtual reality is a real technical challenge for game developers and creators of virtual environments. Visual problems are better known, including the problem of synchronizing the movements in the game with those of the head. An offset that can quickly give you nausea. If it is not likely to make people sick in case of desynchronization, the sound environment also contributes greatly to the immersive side of this technology.

The main difficulty encountered on the auditory level is to generate sound effects correctly located in the space around the player. To do this, the sound should be slightly shifted between the two headphones, a technique already widely used by the industry of cinema. However, noises are much less predictable in a virtual environment. To achieve this, the creators of immersive worlds must provide a large number of models taking into account the location of the sound, but also the type of sound generated by each type of object.

Creating realistic sounds required hours of calculations

The situation is all the more complex as the listener moves in the environment, changing the direction and volume of the sound to be generated. With the current methods, each sound model requires a lot of computation power, monopolizing a cluster of servers for several hours. In front of such an obstacle, the immersive side remains rather limited for the sound effects in the virtual reality. Based on this, a group of researchers from Stanford University in the United States worked on a new method to generate these models much faster. In a recently published article, they describe their new algorithm called KleinPAT.

So far, the generation of sound 3D models was based on the work of Hermann von Helmholtz, a scientist of the XIX th century. It relies on the Finite Element Border or BEM method, a very slow process. Stanford researchers have succeeded in speeding up the calculations for the sound models by changing their approach, by simply avoiding the Helmholtz equation and the BEM method. Researchers believe that this new method speeds up the generation of models, making the process hundreds, even thousands of times faster.