Low Power Streaming of Sensor Data Using Gray Code-Based Approximate Communication
Somayeh Sadeghi-Kohan, Sybille Hellebrand, Hans-Joachim Wunderlich · 2023
The streaming of sensor data has to comply with application specific requirements concerning performance and power dissipation. As it is inherently robust against limited inaccuracies, approximate communication is an ideal solution for optimizing performance and power. While previous approaches have mostly focused on one of the two aspects, the novel scheme presented in this paper targets both performance and power by rigorously exploiting the opportunities of approximation within predefined limits. Starting with base-delta encoding and a given error threshold, the stream of delta values is approximated, such that bus encoding using a Gray code is possible with a minimum number of transitions. The experimental data show that the proposed method maintains the performance gain of a traditional base-delta encoding while further reducing the number of transitions down to a few percent of the original data stream.