Time domain characterization of window length and type on moving variance signal features

Moving variance is a signal used in many applications such as propeller speed identification, EMG artifact extraction, and lane-change recognition. It is also used in the preparation of numerous physiological signals for classification and measurement applications because it can be quickly calculated and reflects the movement or power contained in the signal. This paper investigates the effect of using a weighted moving variance. This paper proposes to quantify the effect of window type and window length on the properties of the moving variance signal. The features used in the time domain characterization are average value and event duration. The parameters were calculated for two signals, one from a respiration band and one from a pressure sensor array. The effect of window type and length was similar on the data from both sensor types. For these measurements, the triangular window was least sensitive to signal fluctuations, whereas the hanning window was the most sensitive. Preliminary figures suggest that an increase window length results in a shorter detected event.