12. Compressed Sensing for High Density Neural Recording

One of the major challenges in large scale electrophysiology recording devices is the volume of data generated. Typically, each electrode samples the neural signal at 30 KHz with 10 bits digital resolution, a typical speed for neural action potentials acquisition. Hence, a 1000 channel neural probe generates data on the order of 300 Mbits per second. For neuroscientists, this presents an enormous problem in both data transmission and data analysis. Recently, as the demand for high density and distributed neural recording devices grows, tackling the problem of data compression and transmission has become extremely urgent. In this chapter, we first summarize a number of techniques used for neural signal compression. We then focus on the recent development on the use of compressed sensing theory to design more efficient high density neural recording circuits.