A technique has been described which minimizes the effects of random noise on quantitative spectroscopic measurements. It is applicable to cases in which a reference spectrum of the analyte compound is available and consists of repeated comparisons of sections of the reference spectrum with corresponding segments of the unknown. It has been demonstrated that the average of these replicate measurements is less affected by noise than is a measurement based on peak height, area, cross correlation, or least-squares fitting. Application to infrared measurements has been demonstrated.
