Carols Pena

Optimal Interleaver Design for Short Blocklength Turbo Codes

Monday, May 22, 2000
11:00 AM
206 Egan

Abstract

This work formulates an interleaver design technique for turbo codes with short blocklength as an optimal solution for a combinatorial optimization problem. The main goal is to find an interleaver which gives the turbo codes low delay and good performance which are typical characteristics in applications like digital voice transmission. The proposed technique can be summarized as an interleaver optimization technique, using simulated annealing algorithm, under two criteria constraints: the minimization of correlation coefficients between extrinsic information and the improvement of its distance spectrum properties. The first criterion is meant to improve the suitability of an iterative decoding process while the second is based on decreasing the code's ``error floor''. Our goal is to apply simulated annealing to optimization algorithm for interleaver design under the criteria mentioned above, such that the optimal solution is obtained via an iterative improvement. In addition, we implement a second simulated annealing approach based on the distance spectrum properties only as a comparison. Computer simulations for bit error rate and frame error rate performance are presented comparing the new technique and some of the best previously proposed interleaver design techniques. The results show that the new technique is slightly superior to the others.

Thesis Committee:
Prof. Masoud Salehi (advisor)
Prof. Bahram Shafai
Prof. Waleed Meleis