Multiple Alignment of Continuous Time Series

Jennifer Listgarten* , Radford M. Neal* , Sam T. Roweis* , and Andrew Emili**
* Department of Computer Science and ** Banting and Best Department of Medical Research , at the University of Toronto.

In Advances in Neural Information Processing Systems 17, MIT Press, Cambridge, MA, 2005.

8 page paper (with colour figures): [1.6 MB pdf], [0.9 MB ps.gz], [15.7 MB ps]

Abstract: Multiple realizations of continuous-valued time series from a stochastic process often contain systematic variations in rate and amplitude. To leverage the information contained in such noisy replicate sets, we need to align them in an appropriate way (for example, to allow the data to be properly combined by adaptive averaging). We present the Continuous Profile Model (CPM), a generative model in which each observed time series is a non-uniformly subsampled version of a single latent trace, to which local rescaling and additive noise are applied. After unsupervised training, the learned trace represents a canonical, high resolution fusion of all the replicates. As well, an alignment in time and scale of each observation to this trace can be found by inference in the model. We apply CPM to successfully align speech signals from multiple speakers and sets of Liquid Chromatography - Mass Spectrometry (LC-MS) proteomic data.


Associated Materials

Code Availability: The Continuous Profile Models (CPM) Matlab Toolbox is now available.

Audio demo of alignment of 10 speech utterances, each from a different speaker (as described in the paper):
original signals, aligned signals (better to save files locally to computer and then to listen rather than listening through your browser).

  • A few slides illustrating the Continuous Profile Model (CPM) as well as figures that could not fit in the paper:

       [html] (works very well on Netscape, reasonably well on Internet Explorer and Firefox)
       [ps.gz 4.7MB] (works better on Unix)
       [pdf 1.4MB] (works better on Microsoft Windows)

  • Also, here is a link to other related papers.

    Back to main page . ----->