This paper introduces a new class of rateless forward error correction(FEC) codes,which are called semi-random Kite(SR-Kite) codes.By introducing fixed patterns to an original purely random sparse parity-check matrix,the proposed code can lower down the error floors.We present a simulation-based greedy optimization algorithm to design the degree distribution of SR-Kite codes for independent Rayleigh fading channels.The performances of SR-Kite codes under maximum likelihood decoding are analyzed for both AWGN and independent Rayleigh fading channels via union bound.Both the analysis and simulation results show that the proposed codes perform well over AWGN and fading channels within a wide range of signal-to-noise-ratios.Moreover,we combine SR-Kite codes with high-order modulation schemes to obtain a continuously variable spectral efficiency in a large range of SNRs. This paper introduces a new class of rateless forward error correction (FEC) codes, which are called semi-random Kite (SR-Kite) codes. By introducing fixed patterns to an original purely random sparse parity-check matrix, the proposed code can lower down the error floors. We present a simulation-based greedy optimization algorithm to design the degree distribution of SR-Kite codes for independent Rayleigh fading channels. The performances of SR-Kite codes under maximum likelihood decoding are analyzed for both AWGN and independent Rayleigh fading channels via union bound.Both the analysis and simulation results show that the proposed codes perform well over AWGN and fading channels within a wide range of signal-to-noise-ratios. Moreover, we combine SR-Kite codes with high-order modulation schemes获得 一 continuously variable spectral efficiency in a large range of SNRs.