ECG data compression for tapeless ambulatory monitors

Abstract

To record two channels of ECG data for 24 h, currently feasible devices require that the ECG be compressed to an average of 200 bits/channel/second or less. The authors describe an ECG data compressor, called TRIM, which meets this goal. Each ECG signal is sampled at 120 Hz with 8-bit resolution and digitally low-pass-filtered to remove components above 40 Hz. TRIM selects a subset of these samples which allows accurate reconstruction of the filtered signal by linear interpolation between them. Significant turning points are selected as the initial output set. Segments of the signal which are bounded by the selected points are partitioned at intermediate points, until the maximum error of linear interpolation between the selected points is below an adaptive threshold. The augmented output set contains about 15 points/s, which are Huffman-encoded using about 10 bits/point.