I have been interested in finding a way to improve performance of speech recognition applications using hardware acceleration. Doing a little research on the web I found some pretty interesting historical information on the development of graphics chips and interfaces such as Direct X. To sum it up, there was a need to create better, faster, richer graphical applications and the primary GPU was insufficient. Not only was it a shared resource that couldn’t guarantee processing but is did not have custom circuitry that efficiently processed graphical moves. Alas companies like NIVIDA were born. Providing boards with custom logic and memory to process 2D and 3D renderings incredibly fast. We have a similar problem with Speech. It is computationally intensive and typically has to share the GPU with all the other applications. The next logical step to creating the “3D” (conversational) of speech is to get some help in hardware. We are at for forefront of the speech application world and doing the same thing that gammers were doing for the graphics cards – pushing the limits! Imagine the difference it makes having a 256MB Multiprocessor graphics card today vs what we had in the late 80’s when Nintendo’s graphics were the best anyone had ever seen. By using hardware to accelerate various aspects of recognizing speech we will be able to transcend current limitations to a world where literally anything is possible. Jump on the bandwagon because its going to happen. Its only a matter of time.