sciquence.similarities.segmental_dtw

sciquence.similarities.segmental_dtw()

Find similarities between two sequences.

Segmental DTW algorithm extends ide of Dynamic Time Warping method, and looks for the best warping path not only on the main diagonal, but also on the other. It facilitates performing not only the comparision of the whole sequences, but also discovering similarities between subsequences of given sequences A and B.

Parameters:
  • A (ndarray (n_samples, n_features)) – First sequence
  • B (ndarray (n_samples, n_features)) – Second sequence
  • min_path_len (int) – Minimal length of path
  • metric (str) – Metric name
Returns:

matchings – List of matching sequences
Return type:

list of list of tuple

See also

dtw()

References

Park A. S. (2006).

Unsupervised Pattern Discovery in Speech: Applications to Word Acquisition and Speaker Segmentation

https://groups.csail.mit.edu/sls/publications/2006/Park_Thesis.pdf