Electrocardiogram data compression using adaptive bit encoding of the discrete Fourier transforms coefficients

Abstract

This study proposes an algorithm for electrocardiogram (ECG) data compression using the conventional discrete Fourier transform. The coefficients are calculated using sine and cosine basis functions instead of complex exponentials, to avoid generation of complex coefficient values. Two well defined strategies are proposed for the choice of the significant coefficients – a fixed strategy based on the selection of a fixed band-limiting frequency, and an adaptive strategy depending on the spectral energy distribution of the signal. The different parameters for the two strategies are empirically selected based on extensive study of a wide variety of ECG data chosen from different databases. The significant coefficients are encoded using a unique adaptive bit assignment scheme to optimise the bit usage. The bit assignment map created to store the bit allocation information is run-length encoded to eliminate further redundancies. For the MIT-BIH arrhythmia database, the proposed technique achieves an average compression ratio of 14.67 for the fixed strategy and 16.58 for the adaptive strategy with excellent reconstruction quality, which is quite comparable to the other reported techniques. The simplicity and low cost infrastructural requirement of the algorithm, makes it suitable for implementation on an embedded platform to be used in mobile devices.