Fast bit error generation for the simulation of MPEG-2 transmissions in wireless systems
P.N. Robert, Ahmed Mohamed Darwish, J.W. Reed · 2003
This paper presents alternate methods to generate bit errors that simulate the effect of a convolutional code on an MPEG-2 sequence. Given the complexity of Viterbi decoders, a fast way to generate bit errors that closely resembles the performance of a convolutional code/Viterbi decoder is desirable in order to minimize the execution time of Monte Carlo simulations. Three methods of generating the bit error sequences are presented in this paper. A simulation of a Viterbi decoder is used as a reference against which the proposed methods are compared. A two-random-variable method is presented that closely matches the performance of the simulation of the Viterbi decoder in both BER and PSNR. A simplification of the two-random-variable method is also presented, where only a single random variable is used, and a deterministic form of the behavior of the convolutional code/Viterbi decoder is used. This single-random-variable method yields results that are similar to the two-random-variable method, with comparable values of BER and PSNR. The statistics of the bit error distribution are sacrificed in exchange for a simplified model and less knowledge of the behavior of the Viterbi decoder. The third method proposed is a simplification of the single-random-variable method, where a single bit error is used to model whole bit error bursts. This method yields inaccurate BER estimates but also PSNR values that are biased but close to the behavior of the convolutional code.