The heart rate variability (HRV) signals can well reflect the effect of exercise on cardiac autonomic nerve function, which can be used as the theoretical index for the development of the best physical exercise program. In this paper, we used multifractal detrended fluctuation analysis (MF-DFA), mass index spectrum and multifractality spectrum to analyze the heart rate variability signals, which have been obtained in these states, such as before, during and after exercise. The purpose is to establish the relationship between the nonlinear parameter of the signals and the autonomic nerve regulation, reveal the physiological mechanism for the effect of circadian rhythm. The result shows that exercise can decrease the correlation and increase the multifractality of heart rate variability signals. In different periods, the diversity of correlation and multifractality among different signals are obviously. The result shows that the correlation of heart rate variability signals was negatively correlated with sympathetic nerve regulation and positively correlated with vagus nerve regulation; the multifractality was positively correlated with sympathetic nerve regulation and negatively correlated with vagus nerve regulation. Further studies show that the function of heart system and the balance of autonomic nerve in the afternoon are more suitable for high-intense physical exercise.