A Fast Hardware Based Hidden Markov Model Predictor for Cognitive Radio

Traditionally Cognitive Radio Networks (CRNs) have been designed to support real-time secondary users (SUs) like VoIP. Hidden Markov Model (HMM) based prediction may provide non-real time access of channels (e.g: Wireless Body Area Network) by the cognitive users in presence of high primary users activity (PUs) that uses sensing information of PU arrival prior to SU transmission. But, HMM model is in general computationally intensive causing high prediction time. In this paper, an advanced model of HMM predictor is designed using reconfigurable FPGA platform that takes the advantage of parallel processing and pipelining architecture. The strength of this work is the design architecture for HMM predictor named as H2M2 engine which is a Hardware Co-simulation (HW-CoSim) based design taking a novel edge trigger based input. This H2M2 engine runs autonomously with minimum interaction with the embedded processor of the cognitive radios without any setup and hold time violations. Thus this predictor is 75% time efficient in comparison with normal HMM predictor making the whole system of cognitive radio transmission energy efficient.