Abstract
A novel metric-based symbol predistortion scheme for peak-to-average power ratio (PAPR) reduction in OFDM systems is proposed in this thesis. The proposed scheme extends or condenses a set of input symbols in the frequency domain by using more reasonable metrics than those of previous literatures, metrics which compute the contribution to the time-domain samples having large magnitudes. The frequency-domain symbols to be extended and condensed are picked as those with the largest and the smallest metrics, respectively. The picked symbols are then extended or condensed by weighting the real and/or the imaginary parts separately. The influence due to the weights, such as the transmitted power, bit error rate, and computational complexity, are also discussed in this thesis. Many parameters in the proposed scheme can be adjusted for different system considerations and thus the proposed scheme has high flexibility. It is also worth noting that the proposed scheme does not need to transmit any side information to the receiver and there is no additional system loading for data detection at the receiver. Computer simulations show that, as compared with previous schemes, the proposed one uses lower average transmitted power to achieve similar PAPR reduction and BER performance with lower computational complexity.