Event-based sampling allows signals to be sampled only when an event occurs. This is particularly advantageous for wireless sensor nodes, as their energy is limited. In the case of send-on-delta sampling with linear prediction, a signal prediction is determined from the last sample value using a Taylor series expansion. The signal is only sampled again when the error in the signal prediction exceeds a certain value delta.

This thesis aims to investigate this approach in terms of reconstruction quality and sampling rate in comparison to classic send on delta sampling.