Attribute-Based Time Series Cross-Correlation Measures

Datasets are usually compared using cross-correlation, often using a moving window to calculate the correlation as a function of time or space. However, signals can be considered as being composed of sinusoids which possess amplitudes, frequencies and phases. All these three attributes can be computed at each point in time, and then used as the basis of a cross-correlation method. By using all the three measures of correlation together, their individual disadvantages can be minimised. By combining these measures with the continuous wavelet transform, information on the correlation as a function of wavelength can be obtained.