Basically, the term “Filtering” is referred to a technique to extract information (signal in this context) from noise contaminated observations (measurements). If the signal and noise spectra are essentially non-overlapping, the design of a frequency domain filter that allows the desired signal to pass while attenuating the unwanted noise would be a possibility. A classical filter could be either low pass, band pass/stop or high pass. However, when the noise and information signals are overlapped in spectrum, then the design of a filter to completely separate the two signals would not be possible. In such a situation the information has to be retrieved through estimation, smoothing or prediction. Figure 2.1 shows a general diagram of an open-loop system (plant) subject to noise contamination at the output end.