Multi-channel ECG data compression using two-stage adaptive vector quantization

Most of the traditional ECG data compression methods are designed for single channel data, and, thus, the correlation among channels is not exploited. Therefore, we propose a new adaptive vector quantization (AVQ) scheme to exploit this correlation for multi-channel ECG data compression. The proposed approach consists of two stages. In the first stage, a traditional AVQ mechanism is applied to each channel independently with its own codebook. Each wining codevector index from each codebook is combined to form a new vector. In the second stage, a similar AVQ mechanism is performed except that this index vector instead of a signal vector is processed and the wining codevector must exactly match with this index vector, i.e., with zero distortion. Therefore, the second stage provides further coding gain without introducing any more signal degradation. With two channels of the MIT/BIH data and three channels of in-house Holter data, our experiment results show that the second stage AVQ allows the increase of compression ratio by 1.26/spl plusmn/0.1 times. In general, the more signal channels involved the more coding gain we can obtain with the proposed approach.