ECG data compression with the Karhunen-Loeve transform

Abstract

The authors analyze a Karhunen-Loeve transform technique for ECG data compression. This transform has been, applied in two different ways: to the entire beat signal and to independent windows (P wave, QRS complex and ST-T complex). The optimum number of coefficients and bits for coding the signal is analyzed for the MIT-BIH Arrythmia database. The data compression performance of both choices are: for the entire beat a mean compression ratio of 12.1 with a mean MSE of 0.3% and for shorter windows a mean compression ratio of 17.21 with a mean value of MSE of 0.44%.