The body area network channel model: A new look at second-order statistics

We adopt an innovative approach, previously applied to first order statistics of body area networks (BANs), across multiple links and multiple subjects, to important second-order statistics of radio propagation for BANs. Here it is demonstrated that for statistics of fade durations, level crossing intervals and non-fade durations, where direct characterizations are made of these statistics, an absolute measure allows us to compare across a spectrum of possible second-order characterizations; ranging from single-parameter for an entire ensemble, through to per-activity, per-subject and per-link based parameterized models. The deficiency of specifying one value (or a few mean values) for either level-crossing rate, average fade duration, or average nonfade duration, with respect to either a median or mean channel gain, is demonstrated with respect to a very large 'open-access' dataset, and particularly 150 hours of 'everyday' on-body link data within that dataset. It is also demonstrated that a 2-parameter lognormal fit, to agglomerate empirical second-order data, is generally simple, accurate and sufficient to characterize fade durations, non-fade durations, and level-crossing intervals.