ECG data compression with vector quantization using k-dimensional tree for fast search

Long-term ECG monitoring demands a bit rate of around 200 bps with acceptable distortion. Vector quantization (VQ) techniques allow one to reach that performance, despite the high computational effort. The most time-consuming stage in the VQ is the search for the best vector in the codebook, and, as a result, this makes real-time implementation infeasible. This paper evaluates the compression of ECG data using VQ with some techniques to optimize the searching for the best vector. The algorithm based on those techniques can be implemented in real-time using low-cost microprocessors. We have evaluated the VQ of the ECG signal and its first-difference using several codebook sizes. The results obtained are good enough to permit a real-time implementation