Current Work

GPUs and Realtime Sound Synthesis

My graduate work involves the use of GPUs for sound synthesis. I am doing my research with William Hsu at San Francisco State University. The final product of this research will be a realtime synthesizer that uses GPUs for computation. This will be freely available to download here when it is available.

In addition to the usual synthesis techniques, we have chosen to leverage the GPUs' power by using compute-expensive techniques for sound synthesis, such as finite difference techniques as suggested by Stefan Bilbao. We have already used the GPU for realtime synthesis using this technique for small grids, 21x21 points and under.

Downloads

Poster for the paper Efficient Finite Difference-based Sound Synthesis using GPUs (below) being presented at SMC2010.

Efficient Finite Difference-based Sound Synthesis using GPUs by Marc Sosnick and William Hsu. This paper was accepted at the 7th Sound and Music Computing Conference in Barcelona, Spain where we will be presenting a poster.

Sound samples referenced in the paper can be found here.

This poster presents the results of the proof-of-concept work that was done to establish the viability of the project, but looks better.

This poster was presented at the San Francisco State University College of Science and Engineering Student Project Showcase on May 14, 2010, and was the second place winner in the Graduate Physical Sciences category.

This poster also presents the results of the proof-of-concept work that was done to establish the viability of the project. The poster was presented at the San Francisco State University Graduate Studies Project Showcase on May 6, 2010.

This Matlab animation attempts to explain how an audio source is taken from a vibrating membrane, based on the Matlab simulation by Bruce Land. Used in the San Francisco State University College of Science and Engineering Student Project Showcase on May 14, 2010. File size is about 620 MB.

Stanford CAMPAIGN Project

The CAMPAIGN project is a collaborative project between Russ Altman's Helix Group at Stanford University, and the San Francisco State University Center for Computing for Life Sciences. Our task is to attempt to modularize clustering algorithms, and optimize them for use on GPUs. The ultimate goal of this work is to provide clustering extensions and speedup of FEATURE.