A 2-D compression of ECG signals is proposed based on context models.R waves are recognized by max absolute value detection and repeated matching.An ECG image is constructed automatically and coding data map is created based on the information of cardiac cycles.Then

1-D discrete wavelet transform and uniform scalar dead zone quantizer are applied to ECG images.Quantized coefficients are decomposed into significance map

sign stream

position of the most significant bit stream and residual bit stream.Based on the coding data map

an adaptive arithmetic coder with different context models is employed for entropy coding of these streams.Two datasets from MIT-BIH arrhythmia database are compressed.With the compression ratio of 20

the percent root mean square difference of 2.93% and 4.31% are achieved

in contrast to 3.26% and 4.8% by JPEG2000 scheme.Results indicate that our method outperforms other compression schemes.