The broader impact/commercial potential of this project lies in the ability of the Extended Analog Computer (EAC) to perform filtering operations not possible with current technology. The low power, near instantaneous filtering of complex analog signals, differentiates the EAC technology from DSP products currently in the marketplace. The development of prosthetics with advanced control characteristics is reaching a computational limit due to the need for neuromuscular waveform recognition and classification in the context of real-time, low power, small form factor computation. EACs promise to surmount these challenges, improving functionality of prosthetics and the quality of life for amputees. In 2010, the global prosthetics market was $3 billion and is expected to reach $4.5 billion by 2017, with myoelectric prosthetics representing a small, but growing fraction of this market. While the myoelectric prosthetics industry represents an initial customer base, this generic computing technology's financial upside can best be estimated by segmenting the multi-billion dollar market for digital signal processing and analog computing devices. The intent is to solve computing problems in niche markets within this broad potential marketplace. This Small Business Technology Transfer Research (STTR) Phase I project is focused on the development of the Extended Analog Computer (EAC) for application to myoelectric prosthetics. New myoelectric interface techniques, such as targeted muscle reinnervation (TMR) are simplifying the use of advanced multi-degree of freedom prosthetics by amputees. However, the dramatic increase in the number and density of electrode sites, and need to implant multi-electrode structures into targeted muscles will increase the signal processing requirements beyond the capacity of traditional mobile digital signal processing (DSP). The EAC is a radical departure from the digital computer, deriving its computational power by taking advantage of the intrinsic solutions to partial differential equations represented as an analog voltage manifold in space. The proposed research aims to implement automatic machine learning/training methods to automatically configure networks of EAC sheets in a radial basis function network (RBFN). The research also explores the effect of the geometry of the input and output points on the sheet to optimize them for the TMR application. Finally, a physical instantiation enabling stand-alone, real-time operation of an EAC-RBFN will be developed. Using recorded data from intramuscular electrode arrays, the performance of the EAC will be tested against standard DSP techniques.
