Abstract： Objective:To develop a training system of Doppler-based venous gas bubble sound discrimination for on-site monitoring of decompression sickness，aiming to solve the training difficulties caused by the lack of actual cases of diving decompression sickness and insufficient Doppler audio data.Methods:Based on a client-server architecture，a training system structure was designed to carry out training at client，and to manage training samples and statistically analyze training records at server. By the algorithm of invoking at the backstage，the Doppler audios of gas bubbles in blood flow were simulated as the training samples. By adjusting the parameters，e.g.，the volume and quantity of gas bubble and the noise level，corresponding to the simulation audios，the success rate of trainees in diagnosing decompression sickness can be quantitatively evaluated.Results:The training system designed in this research functioned as a front-end platform of Doppler gas bubble sound discrimination training，which could quantitatively evaluate training effect，offer massive simulation training for multiple trainees simultaneously，and manage the training data.Conclusion:This training system could be applied to develop the capabilities of Doppler gas bubble sound discrimination and on-site diagnosis and monitoring of decompression sickness of the medical support personnel in the submarine rescue troops at grass-roots level. It can also provide a basis for the selection of on-site diving medical support personnel.