Variance-based signal conditioning technique: Comparison to a wavelet-based technique to improve spike detection in multiunit intrafascicular recordings

Detection of single unit action potentials (APs) from peripheral nerve recordings is complicated by low signal-to-noise ratio (SNR) due to the activity of nearby muscles, interference from more distant nerve fibers, and thermal noise from the neural interface. In this study, we propose a novel signal conditioning technique for multiunit signals (i.e. a signal comprised of multiple units coming from different nerve fibers), based on the variance to be applied prior to detection of APs. The proposed technique was tested on experimental and simulated intrafascicular recordings; and was compared to a wavelet-based conditioning (also applied before AP detection). The outputs of both conditioning schemes were sent to an AP detection algorithm that used a simple threshold (equal to the standard deviation of the signal). The overall performance of the detection phase was superior when using the wavelet-based conditioned signal especially for SNR ≤ 2 dB. However, when using the variance-based conditioned signal, the AP detection phase resulted in lower number of false positives for SNR > 2 dB. The novel variance-based method improves the SNR by attenuating the background noise between APs and can be applied as pre-conditioning processing for AP detection.