This paper investigates the minimisation of the computational cost of providing Immersive Voice Communication (IVC) for mobile clients attached to a Distributed Virtual Environment (DVE). When experiencing IVC, each listening avatar in the DVE, receives a mix of the surrounding avatars' voices, all rendered according to their respective positions in the virtual world. We propose to deliver our mobile IVC service using a type of server-based network architecture to perform voice spatialisation function and mixing for each mobile client. This model will minimise both the access bandwidth and processing power requirements of the mobile devices. We propose a model for computation reduction in servers that re-uses the result of audio spatialisation for several clients by allowing certain level of voice localisation error. Using Linear Programming (LP), we have developed two optimisation formulations to examine different scenarios of applying the computation reduction scheme, with different avatar densities and under varying rendering accuracy constraints. The insights gained from our analytical results should benefit future IVC service providers for mobile DVE.

No abstract

This paper compares three possible rendering models for the provision of Immersive Voice Communications (IVCs) in Distributed Virtual Environments (DVEs) such as multiplayer online games. The common aim of these three rendering models is to create a personalised auditory scene for each listening avatar, consisting of a mix of the surrounding avatars' voices, positioned according to their positions in the virtual world. The first two rendering models are based on amplitude panning localisation and HRTF-based binaural localisation respectively. The computation cost of the latter is deemed too large to meet the identified processing power constraints. A computation reuse scheme was introduced in the third rendering model which, as shown in our simulation results, reduces significantly the computational cost of providing IVC using HRTF-based binaural localisation. Disciplines Physical Sciences and Mathematics

Abstract. This paper examines how earnings benchmarks of firms listed in Mainland China may lead to earnings management through use of discretionary accruals. We compared discretionary accrual levels of small profit firms and small loss firms before and after controlling for the effect of two published benchmarks. We find that firms beating or missing both earnings level benchmark and earnings change benchmark tend to save discretionary accruals, while firms missing either of these benchmarks might employ extra discretionary accruals to meet the other earnings benchmark. Small loss firms reporting a bigger earnings increase might have driven accounting accruals up, and as a result small loss firms as a whole appear to have a high level of discretionary accruals not significantly different from small profit firms. Our findings provide direct evidence for earnings management of listed firms, and also give a reasonable explanation for the "kink" in earnings distribution.

The interactive experiences of players in networked games can be enhanced with the provision of an Immersive Voice Communication Service. Game players are immersed in their voice communication experience as they exchange live voice streams which are rendered in real-time with directional and distance cues corresponding to the users' positions in the virtual game world. In particular, we propose a Mobile Immersive Communication Environment (MICE) which targets mobile game players using platforms such as Sony PSP and Nintendo DS. A computation reduction scheme was proposed in our previous work for the scalable delivery of MICE from a central server. On the basis of that computation reduction scheme, this paper identifies what factors, and to what extent, affect the unacceptable voice rendering error incurred when providing MICE. In the first experimental scenario, we investigate the level of unacceptable voice rendering error incurred in MICE for different avatar densities or avatar population sizes, with a fixed level of processing limit. In the second experimental scenario, we studied the level of unacceptable voice rendering error incurred in MICE for different processing resource limits, with a fixed avatar population size or avatar density. Our findings provide important insights into the planning and dimensioning of processing resources for the support of MICE, with due considerations to the impact on the unacceptable voice rendering error incurred.