Sparse sensing matrix based compressed sensing in low-power ECG sensor nodes

Compressed Sensing (CS) is an acquisition technique able to reduce the operating cost (e.g., energy requirements) of a signal processing system thanks to its capability of simultaneously sampling and compressing an input waveform. Here we focus on Electrocardiogram (ECG) signals acquired by means of a custom designed acquisition board that exploits CS as early-digital compression stage. We show that when CS acquisition sequences are sparse ternary, i.e., with symbols {-1, 0, +1} and designed to maximize their rakeness, it is possible to achieve a reduction in the energy required for ECG signal compression by a factor between 25 and 30 with respect to the standard acquisition with independent and identically distributed random sequences.