Adaptive detection of action potentials using ultra low-power CMOS circuits

Abstract

We present ultra low-power CMOS analog circuits for automatic detection of action potentials (APs). The proposed detection strategy locates AP waveforms and completely preserves their integrity. An adaptive threshold is implemented using a local time-averaging filter presenting a large time constant. The filter uses very small transconductances implemented by means of dedicated circuit techniques and subthreshold operation of MOS transistors. Also, a compact voltage squarer pre-processor is introduced to emphasize neural APs prior to detection. The proposed circuits were implemented in a CMOS 0.18-mum process and achieve ultra low-power consumption. Both circuits have been validated in simulations with synthetic neural waveforms. The adaptive threshold circuit dissipates only 27.2 nW, whereas the voltage squarer dissipates 76.7 nW.