Abstract: Sinusoidal transform is separated into audio analysis and synthesis phases. Referring to the ideas of progressive refinement of geometry/image and scalable video, head-related transfer function (HRTF) filtering is naturally integrated into a progressive resynthesis process and MPEG psychoacoustic model 1 is used to guide the selection of matching pairs before transmission/synthesis. The implementation of this scheme in a distributed virtual environment has quite a few advantages, including scalability, direction flexibility and user-adapted HRTF etc.