Waveform fitting is a key point in data processing of airborne laser bathymetry (ALB), which can provide the foundation data for water depth calculation, submarine sediment classification and water turbidity analysis. Traditional waveform fitting algorithms are often disturbed by noise. In addition, the problem of the present algorithms is that the fitting of complex waveform is not accurate. Therefore, a new waveform fitting algorithm for ALB based on layered heterogeneous model was proposed in this paper. According to the corresponding characteristics of different components of waveform, the ALB waveforms were fitted to a combination of three functions: a Gaussian function for the water surface contribution, a B-spline function for the water bottom contribution, and a Double-exponential function to fit the water column contribution. The performance of the proposed fitting model was verified by the measured data from the South China Sea, compared with three classical waveform processing algorithms: the Double-Gaussian, Generalized-Gaussian, and Richardson-Lucy (RL) deconvolution. The experimental results demonstrate that the proposed fitting model performs best in terms of waveform fitting accuracy and efficiency. The average running time T of the proposed fitting model is 0.019 4 s, saving 0.328 6 s than RL deconvolution. The proposed fitting model performs significantly better than the Double-Gaussian algorithm by reducing 65.11%, 2.83%, 1.01% and 86.61% of their average root mean square error(RMSE), average coefficient of determination(R2), average correlation coefficient(CORR) and average correlation coefficient standard deviation(STD), respectively. The proposed fitting model has the great robustness and can effectively meet the technical requirements of the scientific research and engineering application for ALB.