Stop Repeating Yourself: Exploitation of Repetitive Signal Patterns to Reduce Communication Load in Sensor Networks

We consider communication in systems where (1) it is desirable to reduce communication load and (2) the signals to be transmitted contain approximately periodic patterns. We propose methods that extract the periodic pattern from the signal in real time and reduce communication between sender and receiver to one sample per cycle period. The proposed methods allow for variable cycle duration and immediately return to full communication if the signal no longer contains the extracted pattern. During reduced communication the receiver provides an immediate real-time signal based on extrapolation and concatenation of the extracted pattern. At the end of each cycle, the receiver additionally reconstructs the signal segment of that cycle by means of time warping, which yields an output that is more accurate but is provided only cyclewise. As a specific example application we consider transmission of quaternion signals from wireless wearable inertial sensors attached to the lower body of a walking human. We quantify how repetitive these signals are and how much they change when the subject simulates pathological gait. We then determine the error between the true signal and the extrapolated and reconstructed signals. The proposed method is found to reduce communication load by up to 80% while remaining sensitive to deviations from the natural variability of human gait.