Computing the Geometric Mean Over Multiple-Access Channels: Error Analysis and Comparisons

The paper deals with the problem of data transmission and function computation of the sensed data in wireless sensor networks, in which multiple sensor nodes transmit their data to one sink node over a wireless multiple-access channel. We focus on the problem of computing the geometric mean at the sink node by merging the data transmission and function computation into one step via an explicit utilization of channel collisions caused by simultaneous transmissions of sensor nodes. The paper provides the analysis of the estimation error and compares the scheme with traditional time division multiple-access based schemes to indicate potential for significant performance gains.