Mathematics
Cross-Correlation Analysis
Quick fact
Cross-correlation is used in GPS receivers to find the exact time delay of satellite signals, allowing your phone to pinpoint your location to within meters.
Why this is interesting
You know how you can recognize a friend's voice in a noisy crowd? Cross-correlation analysis works similarly—it can find a known pattern hidden inside a long, messy signal. But how does it 'listen' for the pattern?
Read the full explanation
Understanding Cross-Correlation Analysis
Imagine you have two songs: one is a short clip, the other is a longer recording. You want to find where in the long recording the clip appears. Cross-correlation does this by sliding the clip across the recording, step by step. At each shift (called a lag), you multiply corresponding values and sum them. When the clip aligns perfectly, the sum is large (a peak); otherwise, it's small. This process produces a new sequence (the cross-correlation function) that shows similarity at every possible lag. The same idea works for any two sequences—like sensor readings, stock prices, or brain waves—to see how and when they relate.
A deeper explanation
Mathematically, cross-correlation for two discrete signals f and g is defined as (f ⋆ g)[n] = Σ f[k] g[k+n], where n is the lag. This operation measures the linear similarity as a function of displacement. The key principle is that the sum of products is largest when the signals match in both shape and phase. Cross-correlation is sensitive to both amplitude and timing; it peaks where the signals are most alike. Normalizing by the energies (like the Pearson correlation) gives a value between -1 and 1, isolating shape similarity from scaling. Cross-correlation matters because it reveals hidden time delays—for example, the lag between an earthquake's P-wave and S-wave arrivals tells seismologists its distance. It also underpins matched filtering, used in radar, sonar, and communication systems to detect known waveforms in noise.