Real-time compression of electrocardiogram using dynamic bit allocation strategy

Electrocardiogram (ECG) compression for patient monitoring is a pre-requisite for efficient utilization of the communication link. Few recent works using delta encoding adopted fixed length symbols were analyzed, and we found that their bits per sample efficiency (BPS) is low, specially in low slope regions of ECG wave. For real-time tele-monitoring, this may lower the link throughput. This paper describes a real-time, lossy ECG compression algorithm based on delta encoding using variable length symbols. For this, the zonal complexity and inter-sample slope was estimated in a fixed length block of 52 samples. Each block was attributed as `complex', `semi-complex' and `plain' by comparing the local measures with global ones, which were continuously updated. The encoder performs a dynamic bit allocation (DBA) algorithm for each block of ECG samples to optimize the BPS efficiency. The algorithms were tested on simulation platform with single lead ECG record from MIT BIH compression test data (cdb) and MIT BIH arrhythmia database (mitdb) at 10 bit quantization level, yielding an average BPS of 1.92 with cdb and 1.91 with mitbdb, and low PRD (1.87 and 2.42 respectively with cdb and mitdb).