Removing spurious signals from glow curves using an optimal Wiener filter.

During readout, the signal of the TLD is occasionally polluted with spurious signals. These most often take the shape of a spike on the glow curve. Often these spikes are only a few milliseconds wide but can have a height that significantly influences the outcome of the dose evaluation. The detection of spikes relies generally on comparing the raw glow curve with a smoothed version of it. A spike is detected when the height of the glow curve exceeds that of the smoothed curve, using criteria based on the absolute and relative differences. The procedure proposed is based on smoothing by an optimal Wiener filter, which is, on its turn, based on Fourier analysis for which numerically very efficient methods are available. Apart from having easy to understand tuning parameters, an attractive bonus is that, with only little additional computational effort, estimates of the position of peak maxima are found from second and third derivatives: a useful feature for glow curve quality control.