In this paper, a new proposed approach using wavelet domain statistical analysis is presented to predict epilepsy seizures. In addition, in seizure prediction systems of wearable devices, while using wireless communication technologies such as Bluetooth, the challenge encountered is the amount of data acquired. Compressive sensing has been investigated as a tool to reduce data rates needed to be transferred and processing time as well. Histograms for segments of various signal states were studied in wavelet domain utilizing different signal processing tools such as differentiator, median filtering, local mean, and local variance estimators after the compressive sensing is applied. Simulation results revealed the possibility to use compressive sensing in epileptic seizure prediction systems. This opens the door for more compact seizure prediction algorithms.