Data Compression of Neural Signaling in an Implantable Microchip

Intracellular neural recording is a technique that has been under development for many years, yet still remains fragile and difficult, and requires immobilization of the neurons being recorded. We are developing a system of implantable silicon chips that will intracellularly record, and locally store, the neural signaling. The principal reason we are doing this is to record from intact, mobile animals. We determined that data storage would be one of the major limiting factors in the implantable electronics. In this paper, I present an algorithm for data compression of intracellular neuron recordings that accomplishes a compression ratio of over 800:1 (0.01 bits/sample at a 2.5Khz sampling rate) with acceptable signal error (approximately 8% PRD). At this rate, we can record from ten intracellular electrodes for over 30 hours. The algorithm uses a combination of run length encoding and template matching.