One of the major drawbacks of OFDM has been the high peak-to-average power ratio (PAPR) that is characteristic of signals with multiple sub-carriers. The high PAPR requires additional back off to achieve linear amplification at the transmitter end which results in inefficient power consumption. This inefficient power consumption is the major impediment in implementing OFDM in portable device. Previous efforts to address this problem have been principally directed at two areas, the reduction of signal PAPR and various methods of achieving linear and efficient power amplification (PA). However, all approaches suffer due to various deficiencies such as complexity, computational time, memory requirements, data rate loss and high distortion. Therefore, this thesis aimed at finding the solution of power control problem by reducing the PAPR of the signal. Two novel techniques are proposed in this thesis. One of the techniques is based on distortion class named as Zero Forcing Peaks (ZFP) and the other is based on Selected Mapping (SLM) technique from probabilistic class. In this thesis a new concept of using Learning Vector Quantization (LVQ) along with SLM has been introduced which is named as LVQ-SLM and can be considered as a major contribution of this thesis. Further two different architectures are proposed for LVQ-SLM and their practicability is investigated by synthesizing these architectures on Field Programmable Gate Arrays FPGA. Result obtained using second technique is quite encouraging. An efficient implementation of SLM is achieved by using LVQ network as it reduces PAPR with minimal computational complexity. The only constrain which has been noted when the number of sub carriers and modulation order increase, the on-chip memory to store prototype vectors and computational requirements also increase. High capacity low power Content addressable memory (CAMs) based hardware can be used to solve this constrain which has also been proposed in this thesis.