Punctured parallel concatenated convolutional lattice codes

Convolutional lattice codes, also known as signal codes, are shown to achieve a good performance but with considerably high decoding complexity due to the unbounded number of resulting trellis states. Recently, an alternative approach that can restrict the trellis states based on the recursive convolutional lattice codes is proposed, and its parallel concatenation extension, referred to as turbo signal codes, is shown to achieve the frame error rate (FER) performance close to the Shannon limit. The main drawback is its reduced information rate associated with parallel concatenation and lack of flexibility in its design. This paper thus proposes a symbol puncturing approach to enhance the flexibility in achievable information rate of the turbo signal codes. Simulation results demonstrate that with the information rate of 2.85 bits per two dimension and the block length of 8200 symbols before puncturing, the FER of 10
^-2
can be achieved within the SNR gap of 0.91 dB from the Shannon limit.