Bufferless Compression of Asynchronously Sampled ECG Signals in Cubic Hermitian Vector Space

• Published in: IEEE transactions on biomedical engineering Vol. 62; no. 12; pp. 2878 - 2887
• Main Authors: Marisa, T., Niederhauser, T., Haeberlin, A., Wildhaber, R. A., Vogel, R., Jacomet, M., Goette, J.
• Format: Journal Article
• Language: English
• Published: United States IEEE 01-12-2015; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Asynchronous sampling; Buffer-less compression; Compressed sensing; Computer architecture; Computer Simulation; Cubic Hermitian basis; Data compression; Data Compression - methods; Electrocardiography; Electrocardiography - methods; Geometry; Interpolation; Signal Processing, Computer-Assisted; Threshold voltage; Transforms; cubic Hermitian basis; Asynchronous sampling; buffer-less compression
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2015.2449901

Asynchronous level crossing sampling analog-to-digital converters (ADCs) are known to be more energy efficient and produce fewer samples than their equidistantly sampling counterparts. However, as the required threshold voltage is lowered, the number of samples and, in turn, the data rate and the energy consumed by the overall system increases. In this paper, we present a cubic Hermitian vector-based technique for online compression of asynchronously sampled electrocardiogram signals. The proposed method is computationally efficient data compression. The algorithm has complexity O(n), thus well suited for asynchronous ADCs. Our algorithm requires no data buffering, maintaining the energy advantage of asynchronous ADCs. The proposed method of compression has a compression ratio of up to 90% with achievable percentage root-mean-square difference ratios as a low as 0.97. The algorithm preserves the superior feature-to-feature timing accuracy of asynchronously sampled signals. These advantages are achieved in a computationally efficient manner since algorithm boundary parameters for the signals are extracted a priori.