A signal is sampled and the samples are divided into windows. One of at least three values is associated with a respective window based on whether the samples in the window are relatively constant, incremental or decremental, to thereby generate an approximation vector. Locations of thresholds in the approximation vector are determined, to generate a threshold position vector. Threshold values are assigned to the threshold positions in the threshold position vector. The threshold values, taken alone or in combination with a threshold position vector, may be used to characterize the signal. A training sequence may be used to adapt the functionality to individual needs.