A weighted averaging technique has been developed to overcome the shortcomings of conventional, unweighted averaging in the case of non-stationary noise. The technique is based on the linear least-mean-square estimate of a periodic signal in a simple model of non-stationary noise. This estimate weights each recorded epoch according to the magnitude of the noise within the epoch. In the estimation of a known signal, the weighted averaging procedure yielded smaller root-mean-square errors in comparison with the normal unweighted average. The weighted averaging procedure offers many advantages over conventional averaging or averaging with automatic gain control preamplifiers.
|Original language||English (US)|
|Number of pages||6|
|Journal||Electroencephalography and Clinical Neurophysiology|
|State||Published - May 1984|
ASJC Scopus subject areas
- Clinical Neurology