December 2012 Archives

A Resumé of Year 1

| No Comments
Dear taxpayers of Denmark. You have been gracious to support me this year and next by Independent post-doc grant No. 11-105218 from Det Frie Forskningsråd. Here is a resumé of my accomplishments for the first year. Most of the publications and submissions are available here.

Journal articles:
  1. B. L. Sturm, "When 'Exact Recovery' is Exact Recovery in Compressed Sensing," IEEE Trans. Signal Processing, submitted Jan. 2012. (Rejected)
  2. B. L. Sturm, "The Tzanetakis music genre dataset: Its faults and the challenges they provide," IEEE Trans. Audio, Speech, Language Signal Process., submitted Jan. 2012. (Rejected)
  3. B. L. Sturm, "Classification Accuracy Is Not Enough: On the Analysis of Music Genre Recognition Systems," submitted Nov. 2012 to Journal of Intelligent Information Systems.
  4. B. L. Sturm, B. Mailhé, and M. D. Plumbley, "On Theorem 10 in ``On Polar Polytopes and the Recovery of Sparse Representations,'' and the ERC of BP and OMP in ``Greed is Good: Algorithmic Results for Sparse Approximation,'' submitted Dec. 2012 to IEEE Trans. Information Theory.
Grant submissions
  1. FP7-ICT-2011C FET OPEN STREP Young Explorers (€1,2 million) submitted Sep. 10, 2012 (waiting to hear)
  2. Programme IFD -- Science 2013 for French-Danish scientific co-operation (€3.000) submitted Nov. 2012 (waiting to hear)
  3. Grant from Center for Digital Music, Queen Mary University of London, awarded June 2012; another proposal submitted Dec. 2012.
Conference papers:
  1. B. L. Sturm, "A Study on Sparse Vector Distributions and Recovery from Compressed Sensing," European Signal Processing Conf., submitted April 2012. (Rejected)
  2. B. L. Sturm, "The Tzanetakis music genre dataset: Its faults and the challenges they provide," Int. Society Music Information Retrieval, submitted April 2012. (Rejected)
  3. B. L. Sturm, "Three revealing experiments in music genre recognition," Proc. Int. Soc. Music Info. Retrieval, submitted Apr. 2012. (Rejected)
  4. B. L. Sturm and P. Noorzad, "On Automatic Music Genre Recognition by Sparse Representation Classification using Auditory Temporal Modulations," CMMR 2012, Queen Mary, University of London, June 2012.
  5. B. L. Sturm, "When 'Exact Recovery' is Exact Recovery in Compressed Sensing," Proc. European Signal Processing Conf., Bucharest, Romania, Aug. 2012.
  6. P. Noorzad and B. L. Sturm, "Regression with sparse approximations of data," Proc. European Signal Processing Conf., Bucharest, Romania, Aug. 2012.
  7. B. L. Sturm and M. G. Christensen, "Comparison of Orthogonal Matching Pursuit Implementations," Proc. European Signal Processing Conf., Bucharest, Romania, Aug. 2012.
  8. B. L. Sturm, "A Survey of Evaluation in Music Genre Recognition," Proc. 10th international workshop on Adaptive Multimedia Retrieval, Copenhagen, Denmark, Oct. 2012.
  9. B. L. Sturm, "An Analysis of the GTZAN Music Genre Dataset.", Proc. ACM Multimedia: Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, Nara, Japan, Nov. 2012.
  10. B. L. Sturm, "Two Systems for Automatic Music Genre Recognition: What Are They Really Recognizing?", Proc. ACM Multimedia: Workshop on Music Information Retrieval with User-Centered and Multimodal Strategies, Nara, Japan, Nov. 2012.
  11. B. L. Sturm "Music genre recognition with risk and rejection," Proc. IEEE Int. Conf. Multimedia Expo, submitted Dec. 2012.
  12. B. L. Sturm "On music genre classification via compressive sampling," Proc. IEEE Int. Conf. Multimedia Expo, submitted Dec. 2012.
Invited seminars:
  1. Departement Elektrotechniek, Catholic University Leuven, Belgium (March 21, 2012)
  2. Intelligent Signal Processing Group, Technical University of Denmark (May 14, 2012)
  3. Speech and Audio Processing Group, Imperial University, UK (May 30, 2012)
  4. Centre for Digital Music of Queen Mary University of London, UK (May 23, 2012)
  5. Sound Software Workshop, Queen Mary University of London, UK (June 18, 2012)
  6. Centre for Mathematical Sciences, Lund University, Sweden (September 14, 2012)
  7. Institut for Kommunikation, Aalborg University, Denmark (Oct. 29, 2012)
  8. Communication Theory Lab, KTH-Royal Institute of Technology, Sweden (Dec. 3, 2012)
Research visits:
  1. Departement Elektrotechniek, Catholic University Leuven, Belgium (March 20-23, 2012)
  2. Centre for Digital Music of Queen Mary University of London, UK (April 16 -- June 28, 2012) Funded in part by EPSRC Platform Grant EP/E045235/1 at the Centre for Digital Music of Queen Mary University of London.
  3. Communication Theory Lab, KTH-Royal Institute of Technology, Sweden (Dec. 3-7, 2012)
  4. Centre for Mathematical Sciences, Lund University, Sweden (Dec. 11-14, 2012)
Conferences:
  1. Technical program chair 2012 Sound and Music Computing, July 11-14.
Godt nytår!
I posted an earlier version of this paper, which I presented a few months ago at the 2012 Adaptive Multimedia Retrieval workshop. I have now finished and submitted the final revision.

The old version had 417 references to MGR, no illustrations, and no tables. The new spiffy version has 467 MGR references, three figures, and three tables! I actually read all the papers a second time, corrected many errors in the old version, and expanded the scope of my analysis. It was, fun. :)

UPDATE! I have now added hyperlinks to the references to make reading much easier. (Thanks for the idea Jort!)

I will be actively correcting and expanding this collection. Let me know what I missed, and what is wrong!
From: J. McDermott and M. D. Hauser, "Nonhuman primates prefer slow tempos but dislike music overall," Cognition, vol. 104, no. 3, pp. 654 - 668, 2007.
We started with an extremely crude and coarse-grained contrast between two pieces of instrumental music: a Russian lullaby played on a flute, and an excerpt of German electronic techno... We measured the average event rate of the lullaby and techno to be 65.26 and 369.23 beats per minute, respectively.
That 370 BPM "German techno track"? "Nobody Gets Out Alive" by Alec Empire --- which is clearly not techno, but Drum & Bass. gosh!

A paper that keeps on giving

| No Comments
From: S. Park, J. Park, and K. Sim, "Optimization system of musical expression for the music genre classification," in Proc. Int. Conf. Control, Auto. Syst., pp. 1644-1648, Oct. 2011.

In this simulation of William Martin Joel 'Piano Man' song 29 other songs were a bit of an experiment to extract. Using a filter with frequency band limited reason to put the instrument used in music to have its own frequency band can be. Person's voice is also statistically 200Hz-3kHz. In this paper, to minimize the voice of a man crushed because his voice had taken to preserve some of the background music focused.
Emphasis is mine. Poor William!
Today, I received my favorite unsolicited doomsday screed from a nutter named Frank. It reads in part (I take an image to preserve its formatting --- and to show it is not written in Comic Sans so I take it seriously):

NullSpace.png Dear Frank,

No worries buddy! The Null Space is a fine place to be. Since it appears you specialize in a different discipline than sparse approximation, I provide the following translations to help you spread your message.

  • Replace "Objective Uncertainty Test" with "Exact Recovery Condition". See the Testament of Tropp (Chapter "Greed is Good").
  • Replace "Hell" with "measurements in \(\mathcal{R}^M\)," and "Earth" with "signal in \(\mathcal{R}^N\)."
  • Define "Closed System" as Earth having mostly zeros, and \(K\) non-zeros. You can say "Earth is Sparse" if you wish.
  • Define "Open System" as the set of all solutions, among which the sparsest one (fewest "mass units" in your words, I think) is what we seek.
  • Replace "polarized First Space" with restricted isometry property (abbreviated RIP -- but don't look too deeply for its non-existent religious reference).
  • Thus the "sequence begins": $$\mathcal{O}(M) = K\log(N/K)$$.
  • NB! Since it appears you are doing most of your eschatology in MATLAB (from the bullet points ">>"), make sure you are using "\" instead of "/". That little mistake has tripped me up a few times in my end times predictions!
  • Also, in accordance with the gospel of "reproducible research", you should make available all your algorithms. It really does have an impact --- plug and play evangelism and such.
Nonetheless, fascinating work Frank! Hope to see you in the Null Space soon.
Hello, and welcome to the Paper of the Day (Po'D): Dynamic Iterative Pursuit. Today's paper is D. Zachariah, S. Chatterjee and M. Jansson, "Dynamic Iterative Pursuit," IEEE Trans. Signal Process., vol. 60, no. 9, pp. 4967-4972, Sep. 2012.

My one-line summary of this work is:
Do not compressively sample each frame of a dynamic signal as if it is unrelated to the previous frames.
The paper proposes some interesting new directions in the recovery of compressively sampled sparse dynamic signals using greedy pursuits. Consider each frame of some film transformed in some way that each is sparse. If the film is of some continuous process, then the transformations will likely exhibit continuity as well, with little change from frame to frame. Then, why not use that information to imbue the recovered frames with continuity? For instance, if transitions are slow, then a pixel that is detected as on in one frame will likely be on in the next frame. That is essentially what the "dynamic iterative pursuit" (DIP) does.

With modeling each support element as a first order AR process (with known decay weight and transition probabilities) with a zero-mean Gaussian innovation signal of known covariance, the authors pose orthogonal matching pursuit in a predictive framework (PrOMP, not to be confused with probabilistic OMP). The results look fantastic as the noise increases, but I must spend much more time in the details to understand what is going on --- and this is a good opportunity to revisit my Kalman filtering notes from a long time ago!

Blog Roll

About this Archive

This page is an archive of entries from December 2012 listed from newest to oldest.

November 2012 is the previous archive.

January 2013 is the next archive.

Find recent content on the main index or look in the archives to find all content.