A proposed data compression technique using the Karhunen-Loève expansion achieves a compression ratio of >12:1 for ECGs/VCGs.
A data compression technique is given which yields a compression ratio slightly better than 12 to 1 for cardiogrmns and is implementable with either a mini- or microcomputer. The technique involves two applications of the discrete Karhunen-Loève expansion, intrinsic components, principal factors, or principal components, all synonyms. The first application reduces the effects of respiration and the various orientations of different patients' hearts, and requires the solution of a 3 × 3 matrix eigenvalue, eigenvector problem for each beat. The second application involves expressing the transformed cardiogram in a Kathunen-Loève series, and requires the solution of the eigenvalue, eigenvector problem for a large matrix. However, the solution, which must be obtained only once for all time, can be performed off line. (The same eigenvectors are used for all patients.) Comparisons are given of the cardiograms reconstructed from the compressed data with the original cardiograms.
Womble et al. (Sat,) studied this question.