Multi-lead ECG Compression Approach using Combination of Linear Transformations

Abstract

This paper presents an approach for linear transformations-based compression of multi-lead electrocardiogram signals. It combines principal component analysis for dimensionality reduction among leads and discrete cosine transform applied on each retained principal component. The transformed data are quantized and encoded using arithmetic encoding. This approach is not bounded to any beat detection algorithm and the logical operations are minimized at the expense of arithmetic ones. The achieved compression ratio ranges from 3 to 7 with acceptable quality of the restored signals.