Testing for White Noise

To test whether a set of samples can be well modeled as white noise, we may compute its sample autocorrelation and verify that it approaches an impulse in the limit as the number of samples becomes large; this is another way of saying that successive noise samples are uncorrelated. Equivalently, we may break the set of samples into successive blocks across time, take the FFT of each block, and average their squared magnitudes; if the resulting average magnitude spectrum is flat, then the set of samples looks like white noise. In the following sections, we will describe these steps in further detail, culminating in Welch's method for noise spectrum analysis, summarized in §5.9.

---

Order a Hardcopy of Spectral Audio Signal Processing

Previous: White Noise
Next: Sample Autocorrelation

---

written by Julius Orion Smith III
Julius Smith's background is in electrical engineering (BS Rice 1975, PhD Stanford 1983). He is presently Professor of Music and Associate Professor (by courtesy) of Electrical Engineering at Stanford's Center for Computer Research in Music and Acoustics (CCRMA), teaching courses and pursuing research related to signal processing applied to music and audio systems. See http://ccrma.stanford.edu/~jos/ for details.