Electrocardiographic data compression using preceding consecutive QRS information

A personal-computer-based electrocardiographic (ECG) noiseless-data compression system has been developed. In addition to using conventional techniques such as a linear predictor and Huffman probability code, the algorithm introduces a new technique to take advantage of the similarity and periodicity of QRS waveforms. The algorithm consists of two cascaded stages. In the first, a linear predictor with three optimal coefficients is used. In the second stage, a portion of the QRS waveform is predicted from previous consecutive QRS complexes to further reduce the bit rate. If no improvement is detected during the second stage, the program automatically employs the first-stage algorithm only. This arrangement ensures the minimum data compression to be equal to that of the conventional technique. A total of nine passages of 15 beats each were processed. Of the total beats, 96% were encoded with both stages of the algorithm and 4% used only the first-stage algorithm. It is concluded that consecutive QRS complexes can predict the subsequent QRS with high accuracy. For noiseless encoding, the average compression ratio was 2.0, and the bit rate was 1015 b/s