IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 7, JULY 2008 1923 and median frequencies are below 10% and are comparable to the ones in [11]; however, the moments are not well preserved and present large variation. A reason for this behavior is that the MMP algorithm works on the time domain, and uses the mean-squared error as the distortion metric; this favors the maintenance of the shape (as can be seen from the good PRD results), but not the spectral features. One possible so- lution for it, that is a promising area for future work, is to add other distortion metrics to 1, taking into account spectral features. Another promising research direction would be to use the proposed algorithm to compress isotonic signals. Since it uses an adaptive dic- tionary, it has the potential to capture and quickly adapt to the nonsta- tionary characteristics presented by such signals. The computational complexity of the MMP algorithm is reasonably high and a simplified analysis concerning this matter is available in [9]. Besides, we have not developed fast algorithms for its implementation yet, since the main goal currently is to identify its potentials and in- vestigate enhancements for improving its performance for a variety of signals, like the EMG. Given these observations, we did not carried out complexity comparisons in the context of this paper, however, this is a main concern for future works. V. CONCLUSION We applied a one-dimensional version of the MMP algorithm to the compression of EMG data. The base algorithm and its extensions, composed of tools that allow better adaptation to smooth sources and more effective use of the dictionary, provided reconstructed signals with high quality. The results obtained for isometric signals were good, outperforming state-of-the-art schemes for EMG compression in terms of PRD × compression ratio. In brief, this paper enforces that the MMP algorithm, due to its universality, is an interesting alternative for bio- logical signal encoding, and can be a viable alternative to compressing other data of the same class, like the EEG. ACKNOWLEDGMENT The authors would like to thank Prof. P. Berger, Universidade de Bras´ ılia, for providing the test signals used in this paper. REFERENCES [1] A. P. Guerrero and C. Mailhes, “On the choice of an electromyogram data compression method,” presented at the 19th Int. Conf.—IEEE/EMBS, Chicago, IL, Nov. 1997. [2] E. Plevin and D. Zazula, “Decomposition of surface EMG signals using non-linear LMS optimisation of higher-order cumulants,” in Proc. 15th IEEE Symp. Comput.-Based Med. Syst., Jun. 2002, p. 49. [3] D. Farina, R. Merletti, and R. M. Enoka, “The extraction of neural strate- gies from the surface EMG,” J. Appl. Physiol., vol. 96, no. 4, pp. 1486– 1495, Apr. 2004. [4] M. Nikolic, “Detailed analysis of clinical electromyography signals emg decomposition: Findings and firing pattern analysis in controls and pa- tients with myopathy and amytrophic lateral sclerosis,” Ph.D. dissertation, Faculty Health Sci., Univ. Copenhagen, Denmark, 2001. [5] T. Hoshino, M. Tomono, R. Furusawa, T. Suzuki, M. Shimojo, and K. Mabuchi, “Development of a motion support system by using an elec- tromyogram,” in Proc. IEEE Int. Conf. Syst., Man, Cybern., The Hague, The Netherlands, Oct. 2004, vol. 5, pp. 4432–4437. [6] P. A. Berger, F. A. O. Nascimento, J. C. Carmo, and A. F. Rocha, “Com- pression of EMG signals with wavelet transform and artificial neural networks,” Inst. Phys. Publishing: Physiol. Meas., vol. 27, no. 6, pp. 457– 465, Jun. 2006. [7] M. B. de Carvalho, E. A. B. da Silva, and W. A. Finamore, “Multidimen- sional signal compression using multiscale recurrent patterns,” Signal Process.: Image Video Coding Beyond Stand., vol. 82, no. 11, pp. 1559– 1580, Nov. 2002. [8] K. Nazarpour, A. R. Sharafat, and S. M. Firoozabadi, “Negentropy analysis of surface electromyogram signal,” in Proc. 13th IEEE Workshop Statist. Signal Process., Bordeaux, France, Jul. 2005, pp. 974–977. [9] E. B. L. Filho, E. A. B. da Silva, M. B. Carvalho, W. S. S. J´ unior, and J. Koiller, “Electrocardiographic signal compression using multiscale re- current patterns,” IEEE Trans. Circuits Syst.—I: Reg. Papers, vol. 52, no. 12, pp. 2739–2753, Dec. 2005. [10] J. A. Norris, K. Englehart, and D. Lovely, “Steady-state and dynamic myoelectric signal compression using embedded zero-tree wavelets,” in Proc. 23rd Annu. EMBS Int. Conf., Oct. 2001, pp. 1879–1882. [11] E. Carotti, J. C. De Martin, R. Merletti, and D. Farina, “Compression of surface EMG signals with algebraic code excited linear prediction,” Med. Eng. Phys., vol. 29, no. 2, pp. 253–258, May 2006. ECG Signal Compression Based on Dc Equalization and Complexity Sorting Eddie B. L. Filho, Nuno M. M. Rodrigues, Member, IEEE, Eduardo A. B. da Silva*, Member, IEEE, S´ ergio M. M. de Faria, Member, IEEE, Vitor M. M. da Silva, and Murilo B. de Carvalho Abstract—In this brief, we present new preprocessing techniques for electrocardiogram signals, namely, dc equalization and complexity sort- ing, which when applied can improve current 2-D compression algo- rithms. The experimental results with signals from the Massachusetts Institute of Technology - Beth Israel Hospital (MIT-BIH) database out- perform the ones from many state-of-the-art schemes described in the literature. Index Terms—Data compression, electrocardiogram (ECG), H.264/ AVC, JPEG2000, preprocessing. I. INTRODUCTION The electrocardiogram (ECG) is a very important tool for medical diagnosis. The need for remote diagnosis may demand for the transmis- sion of the complete exam, which may be composed by as much as 12 derivations, over bandwidth-restricted communication networks. Fur- thermore, the need for databases of ECG exams from various patients, aiming at pathology development analysis or comparative diagnosis, makes efficient storage also an important issue. Therefore, both ECG storage and transmission demand suitable compression schemes, which Manuscript received June 13, 2005; revised August 19, 2007. Asterisk indicates corresponding author. E. B. L. Filho is with the Centro de Ciˆ encia, Tecnologia e Inovac ¸˜ ao do P´ olo Industrial de Manaus, Manaus-AM 69057-040, Brazil (e-mail: eddie@ ctpim.org.br). N. M. M. Rodrigues and S. M. M. de Faria are with the Instituto de Telecomunicac ¸˜ oes and ESTG. Inst. Polit´ ecnico Leiria, Morro do Lena - Alto do Vieiro, 2411-901 Leiria, Portugal (e-mail: nuno.rodrigues@co.it.pt; sergio. faria@co.it.pt). ∗ E. A. B. da Silva is with the Universidade Federal do Rio de Janeiro, Rio de Janeiro-RJ 21941-972, Brazil (e-mail: eduardo@lps.ufrj.br). V. M. M. da Silva is with the Instituto de Telecomunicac ¸˜ oes and Dep. Eng. Electrot´ ecnica e de Computadores, P´ olo II - Universidade de Coimbra, 3030-290 Coimbra, Portugal (e-mail: vitor.silva@co.it.pt). M. B. de Carvalho is with TEC/CTC, Universidade Federal Fluminense, Niter´ oi, RJ 24210-240, Brazil (e-mail: murilo@telecom.uff.br). Digital Object Identifier 10.1109/TBME.2008.919880 0018-9294/$25.00 © 2008 IEEE