Biomedical signal compression with time- and subject-adaptive dictionary for wearable devices

Abstract

Wearable devices allow the seamless and inexpensive gathering of biomedical signals such as electrocardiograms (ECG), photoplethysmograms (PPG), and respiration traces (RESP). They are battery operated and resource constrained, and as such need dedicated algorithms to optimally manage energy and memory. In this work, we design SAM, a Subject-Adaptive (lossy) coMpression technique for physiological quasi-periodic signals. It achieves a substantial reduction in their data volume, allowing efficient storage and transmission, and thus helping extend the devices' battery life. SAM is based upon a subject-adaptive dictionary, which is learned and refined at runtime exploiting the time-adaptive self-organizing map (TASOM) unsupervised learning algorithm. Quantitative results show the superiority of our scheme against state-of-the-art techniques: compression ratios of up to 35-, 70- and 180-fold are generally achievable respectively for PPG, ECG and RESP signals, while reconstruction errors (RMSE) remain within 2% and 7% and the input signal morphology is preserved

Overcoming Bandwidth Limitations in Wireless Sensor Networks by Exploitation of Cyclic Signal Patterns: An Event-triggered Learning Approach

“…[27] (lossless entropy compression, LEC), ref. [28] (grouping and amplitude scaling), and [29] (time-and subject-adaptive dictionary). All these methods introduce delays of at least 1 s. In consequence, they are not suitable for real-time tracking.…”