Optimizing signal coding in neural interface system-on-a-chip modules

With the recent technology advances in multichannel microprobe fabrication for neural recording, many limitations still encumber the associated processing and communication capabilities. Consequently, optimizing the information transfer from multiunit neural recording devices is strongly motivated to better understand the underlying mechanisms of neural information processing in the central nervous system in real time. We recently developed a framework aimed at processing neural signals with optimized computational complexity. In this work, we propose an associated coding solution aimed at implementing fast and efficient neural interface system-n-a-chip (SOC) modules capable of exchanging neural information without compromising issues of communication bandwidth and signal fidelity. Preliminary results demonstrate that neural information processing and on-chip coding offers tremendous savings in communication costs compared to raw data transmission for off-chip analysis. Moreover, we demonstrate that bit rates in the order of 1 bit/sample can be easily achieved with a 32 channel device and 25 kHz sampling rate. Performance illustrations and experimental neural data examples are described in details.