Robust graph-based denoising for cardiac acceleration signals

This paper proposes an efficient approach for denoising heart vibration signals captured by a 3D accelerometer in an implantable device located in the gastric fundus, aimed at improving heart failure monitoring. The approach leverages the inherent consistency (i.e. pseudo-periodicity) of heart vibration signals across cardiac cycles, reformulating the denoising problem as the inference of a low-rank matrix under the assumption of signal smoothness on graph structures associated with the target signals. Both the group sparsity and the total graph variation concepts are employed to describe the aforementioned assumptions. The effectiveness of this method, compared to standard denoising techniques, is confirmed by using real 3D accelerometer signals acquired from seven pigs with and without heart failure.