On CS Image Reconstruction Using LDPC Code Over Radio Mobile Channel

Abstract

In this work, the effect of additive white Gaussian noise and fading channel on the compressed sensing or compressive sampling (CS) image reconstruction process are demonstrated. First, the work suggests encoding of the sensed samples by low density parity check code (LDPC) before transmission. It is well known that longer is the length of the LDPC codes better (lower) is the bit error rate (BER) performance. Thus to improve CS reconstruction a method to construct a larger length but 4 cycle free irregular LDPC code structure is also proposed. The code construction is based on the LDPC code in IEEE WiMAX standard. The proposed CS-LDPC structure is then extended for $$4^n$$ -QAM to find an optimal set of thresholds by minimizing BER (equivalently symbol error rate) using differential evolution (DE). The algorithm works on the log likelihood ratio values obtained by LDPC decoding. Extensive simulation results show the efficacy of the use of LDPC codes and the trade-off in code rate and measurements on reconstruction quality. Improved performance with the proposed DE based demodulation is also demonstrated.