A μW-level Multi-channel Calibration-free Spike Detector with High Accuracy based on Stationary Wavelet Transform and Teager Energy Operators

A prevailing trend in brain-computer interface (BCI) systems is expanding the number of recording channels to the realm of thousands. The massive quantities of raw data impose substantial demands on the data transmission bandwidth, leading to increased power consumption and thermal dissipation within implanted systems. In response to these challenges, this article proposes a stationary wavelet based Teager energy operator (SWTTEO) spike detection algorithm with adaptive thresholding, which dramatically compresses the data bandwidth. The algorithm facilitates spike detection with an accuracy exceeding 97% even under circumstances with the noise level as high as 0.2. The lifting scheme of the db3 wavelet is employed to reduce the hardware resource. The proposed spike detector for multi-channel neural interface is implemented in 65-nm and 180-nm CMOS technologies in a channel-interleaved architecture. The optimized 65-nm implementation consumes a power of 1.07 μW and an area of 4048 μm
²
for each channel.