SLVA decoding with unequal error protection

Data with varied reliability requirement, for example, digitally encoded speech, need a coding scheme that can provide different levels of protection. An error in the most significant bits can be more damaging than an error in the least significant bits. Therefore, channel coding with unequal error protection (UEP) is needed. Rate-compatible punctured convolutional (RCPC) codes can be used to apply unequal error protection. In this paper, the serial list Viterbi algorithm (SLVA) decoding with UEP is studied. We observed that the error performance of SLVA decoding depends strongly on the puncturing provided to the least significant bits. Limiting the SLVA search window to cover the most important bits, the SLVA is able to correct the errors occurring in the most significant bits more efficiently. The effect of location of the most important bits either in the beginning or in the end the sequence was evaluated. When the most important bits are located at the beginning of the information sequence, the SLVA search window should be expanded by few bits of the area where these bits are located in order to achieve the best performance with SLVA decoding. The effect of CRC code for error detection was also studied. The length of the CRC code is a crucial factor, since the probability of undetected errors should be kept low in order to. achieve the best performance with a limited SLVA search window.