Advanced software tools for parametric identification based on quantized data

Parametric estimation of signals, based on quantized data, is often carried out by means of least squares (LS) or averaging techniques. Such an approach often leads to optimal performance, resulting in almost unbiased estimators when the quantization error can approximately be modeled as an additive white Gaussian noise, or when other additive white Gaussian noise sources are larger than the quantization error. When such hypotheses are not satisfied, however, averaging may produce suboptimal, and biased estimators. In such a case, maximum likelihood or quantile based identification techniques can be shown to lead to more performing estimators, mostly unbiased and with a lower mean square error than that of an LS estimator. A software tool is presented, capable of estimating a DC level, a DC level corrupted by Additive White Gaussian Noise (AWGN), and sinewave parameters when the frequency is known, using data quantized by a nonuniform ADC.