This section describes the choice of parameters used in the project.
THE ENCODER:
The enocoder parameters are the length M of the sequences, as stated in the BASIC SCHEME, and the error-control code used. The length M is selected to be 4, i.e., X data are separated into groups of length 4 to be Huffman coded. The reason for this selection is that shorter lengths do not provide sufficently equally likely bit planes; while longer M lengths cause large errors. The reason for the introduction of large errors with long M lengths is because even one change in any of the M bits in the sequence changes the Huffman code. Hence, keeping the length M short provides lower error rates, and hence, better decoding at the decoder.
For the error-control code, BCH codes of (n,k) with n=15 are used. The k parameter is varied, providing the bit-rate used in the transmission. The syndromes are computed using the parity-check matrices of the BCH (15,k) codes. The k values used are 1,5,7 and 11.
THE DECODER:
The decoding is done as described in the BASIC SCHEME section. All of the codewords corresponding to the received syndrome are computed using the parity-check matrix, and the one closest in Hamming distance to the best estimate is selected as the correct transmitted codeword. The best estimate is computed using the conditional probability given the received sequence, given Y, the noisy data at the decoder, and the probability of error, p.
THE NOISE:
The noise is i.i.d., Gaussian, and is independent of the correlated data sets X and Y. The variances are selected such that p (the probability of error in the binary symetric channel model) is between .05 and .10.
THE PROBABILITY OF OCCURRENCE FOR 0's and 1's:
The probability of occurence for 0 (or conversely for 1) is varied from 0.50 to 0.90. The length of the data set used is 60000.
THE MAXIMUM ACCEPTABLE ERROR RATE:
The minimum acceptable error rate is selected as 0.005, and is defined as the number of 0's and 1's that the original data set X and the decoded data differ by.
ABSTRACT
INTRODUCTION
PRIOR WORK
BASIC SCHEME
METHODOLOGY
RESULTS
CONCLUSIONS
REFERENCES