ECG compression using PCA with polynomial estimation of higher order coefficients

An ECG compression technique which achieves very low bit rates of is 12bps is presented. Principal Component Analysis (-PCA) is used prior to identifying non-linear relationships between the orthonormal eigenvectors. Polynomials are used to approximate higher index transformed coefficients from lower ones. Reconstruction is similar to that used in the Karhuneun-Loeve transform (KLT) technique but some of the coefficients have been estimated rather than stored explicitly. Exceptionally low bit rates have been achieved with good reconstruction appearance. For very low bit rates, improvements are consistently seen over KLT and however further work is required concentrating on generating for-clinically acceptable performance on all examples possibly through relaxing the bit rate requirements slightly to achieve lower reconstruction errors.