Portal-based Sound Propagation for First-Person Computer Games Cameron Foale University of Ballarat Mt. Helen, Australia c.foale@ballarat.edu.au Peter Vamplew University of Ballarat Mt. Helen, Australia p.vamplew@ballarat.edu.au ABSTRACT First-person computer games are a popular modern video game genre. A new method is proposed, the Directional Propagation Cache, that takes advantage of the very com- mon portal spatial subdivision method to accelerate envi- ronmental acoustics simulation for first-person games, by caching sound propagation information between portals. 1. INTRODUCTION Sound is vitally important in creating a sense of player im- mersion in first-person computer games. Sound contributes strongly to the sense of presence, or of “being in the game” [43, 6]. Sound cues are used by players to determine the in-game locations of non-visible events and objects, and to reinforce visual information, helping players to gauge the distance of visible objects [37]. Further, sound is an impor- tant dimension for artistic input in the design of games. Acoustics auralization is the process of simulating how a sound would be heard if it was placed in an environment. Auralization requires calculating how a sound propagates from a sound source to a listener. Sound propagation from a sound source to a listener is described by an acoustic transfer function. Existing games use relatively unsophisticated acoustic trans- fer functions, such as simple low-pass filters and generic de- lay and reverb effects. These functions are computation- ally cheap. However, the resulting sound environment is only loosely based on how sound would actually propagate around the environment. Valve’s Half-life 2 [42] presents a more sophisticated ap- proach, interpolating the parameters of delay and reverb ef- fects between artist-specified settings, or soundscapes, based on the current environment. For example, a sound designer creates a large reverberant room soundscape, and a small reverberant room soundscape, and the audio engine interpo- lates between these depending on the size of the room con- taining the player. This method can produce highly stylised artistic effects, however there is no attempt to model sound propagation, and the effects do not depend on the direction and location of the sound source or listener. The most common acoustic transfer function in architectural acoustics is the impulse response. An impulse response de- scribes how a single “spike” of sound pressure (an impulse ) at the source position arrives, over time, at the listener po- sition. The simplest way to conceptualise the impulse re- sponse is to make a loud, sharp noise (such as a hand clap or bursting balloon) in a real environment - the combina- tion of echoes heard is an impulse response, unique to the source location and the position of the listener. A recorded or generated impulse response can be mathematically ap- plied to any sound signal using convolution, to “place” the sound at the location that the impulse was created, however convolution is a computationally expensive process without dedicated hardware. Simulating an accurate impulse response in real-time is too computationally expensive for computer games. However, for other aspects of computer games, performing calculations off-line and then applying the results in real-time is common, such as offline calculation of global illumination offline for real-time lighting and shading [20, 25], and offline spatial subdivision and potentially visible set calculations for real- time rendering [1]. We suggest that there is a large demand for higher quality sound simulation in computer games, and there is scope to use off-line computations to improve real- time auralization. 1.1 Structure This paper is structured as follows: • Sections 2-5 provide background on auralization and sound phenomena • Sections 6-8 describe existing autralization methods • Sections 9 and 10 describe the new method, and means for evaluation 1.2 Assumptions The following simplifying assumptions and approximations are common in architectural acoustics literature, and are also made for this paper: