Reliable communications in wireless body area networks

This thesis investigates the problem of reliability in wireless body area networks(WBANs). WBANs constitute a class of wireless networks that are composed of miniaturized wearable or implantable nodes inside or in the vicinity of human body, with diverse applications in medicine, personal care, and entertainment, or more broadly Internet of Things (IoT). Since WBANs can potentially convey sensitive health-related signals,high reliability of the wireless links is a necessity. Typically, reliability of communications can be increased by leveraging node resources such as energy or computational complexity. Nonetheless strict limitation of such resources in small battery-operated WBAN nodes to overcome severe channel conditions unique to WBANs pose a significant challenge that demand high transmission effciency with low complexity. A key idea in this thesis is that achieving an optimal transmission effciency can relax the constraint on node resources such that the node's expected lifetime is increased and it can allocate more energy and computational power to countering harsh noise and fading conditions. The thesis studies different types of WBANs and propose novel techniques across different layers of the communication protocol to achieve high transmission effciency. The proposed methods rely on statistical signal processing, adaptation, error control coding, and optimization. However, maintaining a low computational demand at the transmitter node is a key requiremen tthat has been considered in all methods. The main problem is addressed in two different parts of this thesis. The first part deals with the problem of optimization of WBANs based on the IEEE 802.15.6 recommendations, which is the state-of-the-art communication protocol for WBANs. Novel MAC-level adaptation and optimization schemes are proposed for impulse-radio ultra-wide band systems and a theoretical framework to achieve the optimal energy-delay trade off for reliable communications is provided. Simulations confirm that the transmission efficiency can be improved by up to a factor of two by link adaptation. Also the energy efficiency is maximized with respect to the frame length and a closed formula is derived. In the second part, a more general system is considered and the problem of optimal transmission efficiency in a typical WBAN/IoT device is addressed. The thesis proposes novel coding techniques based on random linear coding (RLC) as well as capacity-achieving low-complexity polar codes to outperform the state-of-the-art error control techniques. Specifically, novel hybrid-ARQ schemes based on systematic polar codes are proposed that are designed for low complexity and short code-length implementations that can achieve about 4 dB gain at low SNR. It is also proposed to leverage the available receiver-side computational power to fix partially corrupted packets without asking for any further redundancy from the transmitter which relies on transmitter-side RLC, combined with receiver-side sparse recovery. This technique can lead to performance gain in point-to-point RLC coded systems and improves transmission efficiency typically by 50% in multicast. This thesis also proposes a novel joint sampling-quantization architecture with nonuniform sampling time and numbers of bits per sample that can represent a segment of a band-limited signal with the smallest number of bits, compared to the state-of-the-art previously known techniques, without performing any transform coding compression. In this way, it is shown that node resources are utilized efficiently and the resources required for quantization, compression, and transmission are saved since fewer bits are produced and transmitted assuming a given segment of signal acquired by the sensor. Simulations show that the proposed sub-Nyquist sampling scheme leads to only 12% of total number of bits compared to the conventional uniform Nyquist sampling.