A comparative study on the estimation of noise standard deviation using DATE and truncation thresholds

D-dimensional amplitude trimmed estimator (DATE) is used for estimating the noise standard deviation in data that can be modelled as a signal in additive white Gaussian noise. Truncation thresholds are a pair of amplitude thresholds that are used for spike detection in extracellular neural recordings. Noise standard deviation is estimated as a byproduct of the computation of the truncation thresholds. Here, standard deviation estimates obtained using DATE and truncation thresholds are compared in realistic simulations of extracellular neural recordings. The results show that noise standard deviation is estimated more accurately and faster with truncation thresholds. These findings are important for developing a suitable method for amplitude thresholding in brain-machine-interfaces.