Figure 8. Indexing earthquake waveforms for fast search.
A MRKD tree method is applied to create a database tree for fast search. (a) Assuming a data set of four seismograms, the mean and RMS variance for each data dimension are calculated for all of the traces; (b) At the top of the tree, the largest variance (0.143) occurs at the sample 864 and the seismograms with amplitudes at this sample less than the mean (0.176) are placed on the left, those larger than the mean are placed on the right. The process is recursively applied at each level until only one seismogram remains on each side.
