A novel adaptive real-time detection algorithm for an area-efficient CMOS spike detector circuit

For implantable neural detection and recording systems, computationally simple methods are preferred over complex algorithms for power efficiency. Amplitude threshold is the simplest and the most widely employed method for neural spike detection. However, this conventional method for setting the threshold is sensitive to the spike firing rate and spike amplitude. In this work, a novel approach is described for a real-time calculation of the detection threshold in a way that is robust to the change in spike firing rate as well as spike amplitude. An algorithm is proposed that adaptively sets the threshold from background activity by approximating the nonlinear energy operator (NEO). The detection method can distinguish neural spikes accurately with firing rate greater than 100 Hz. The spike detector circuit dissipates 5.1 μW of power per channel and occupies 0.018 mm2 of layout area when implemented in 0.18 μm CMOS process.