It is shown that the discrete cosine transform (DCT) of a given set of elec
trocardiographic (ECG) cycles can be approximated to a set of narrow band s
ignals. These signals are estimated by performing singular value decomposit
ion (SVD) on a prediction matrix made up of the DCT coefficients. The QRS p
ulses are then separated out in the DCT domain from a reduced rank approxim
ation of the prediction matrix. The size of this reduced rank matrix is obt
ained by means of an optimally established orthogonality relationship betwe
en two complementary subspaces derived from the signal space of DCT coeffic
ients. (C) 1999 Elsevier Science B.V. All rights reserved.