• ICDM2011 paper and presentation.
• Slide to show all figures in higher resolution.
• Matlab source code to run all experiments is here.

• An interwoven calls between Elf Owl (mp3) and Pied-billed Grebe (mp3) from AZFO Bird Sounds Library. 
• The top time series shows an audio of the interwoven calls we created.  
• The bottom time series shows the time series converted by applying Mel-frequency cepstrum (MFCC) technique.
• You can download MFCC time series here. 
• Vdo shows the clustering created from our algorithm is here.
• Matlab code for reproduction is here.

• A poem "The Bells" by Edgar Allan Poe.
• The original audio file (mp3) is 4-minute long downloaded from LibriVox collection of Internet Archive.
• We do clustering on last 30-second excerpt shown in the left figure. The middle figure shows that same things, but text is grouped by clusters. 
• You can see the result of clustering algorithm in this vdo.
• This is the code for reproducing the experiment.
• Mirror: 30-sec poem in audio format (mp3), time series transformed by MFCC technique (txt).
• Mirror: 4-min poem in audio format (mp3), time series transformed by MFCC technique (txt). Note that the introduction and ending parts are removed. 

• Winding dataset contains seven dimensions. We provided a mirror of the whole dataset here. 
• The top figure shows the time series in dimension U1 (the first dimension) of the Winding dataset; subsequences in same color are in the same cluster.
• The middle figure shows subsequences in five clusters from seven clusters (total) of U1 dimension. 
• The bottom figure shows the clustering results of the Winding time series in the forth dimension.
• A vdo to show clustering result from our algorithm on all seven dimensions is here.
• This is the code for reproducing the experiment.

• A heart beat time series from PhysioBank ATM. (Dataset: ANSI/AAAMI EC13, Record: aami3c, Signals: ECG)
• The mirror of the data is provided here.
• Clustering result generated by our algorithm is here. Please consider at step 7 and 8.

• A heart beat time series from PhysioBank ATM. (Dataset: Challenge 2010 Training Set A, Record: a05, Signals: ECG V) 
• We provided the mirror of the time series here.
• Clustering result generated by our algorithm is here. Please consider at step 8 and 9.

• The top time series shows the temperature of Jan 21st, 2006 (recorded in the file dir.20060121_001_ceps.txt.
• Whole dataset contains temperatures of 56 days and 90 days is here.
• The vdo to show clustering is here.