Service robots are gradually applied to emotional accompaniment and emotional healing. The work aims to explore the emotional healing mechanism of cuteness morphology robots, and to study the elements of cuteness morphology features of robots that affect users' emotional experience. The semantics of cuteness morphology were analyzed and a sample of cuteness morphology robots was selected to conduct human-robot interaction experiments. Users' facial electromyographic signals (fEMG), electrodermal signals (EDA), and heart rate variability signals (HRV) during stress, calmness, and human-robot interaction were measured, and subjects were invited to fill out the subjective questionnaire of the robot's cute experience after the interaction. In conclusion, fEMG indicators show that interacting with the cuteness morphology robot after a stressful stimulus can effectively arouse positive emotions. The higher the robot cuteness score, the higher the user's positive emotion score. The EDA indicators show that the subject's emotional arousal degree is also higher. The same robot form with cute expression makes the SDNN indicators in the HRV signal of the subject significantly higher. The same robot form with the tactile interaction of the soft material makes the SDNN and VLF indicators in the HRV signal significantly higher; The same action interaction with the robot of higher cuteness makes the subjects’ EMG signal, EDA signal, and HRV signal indicators significantly higher. The SDNN and VLF indexes in HRV signals are significantly increased by the same robot form. The EMG signal, EDA signal, and HRV signal indexes are significantly increased by the same robot form in the same action interaction. Finally, physiological indicators that have correlation with cute experience are selected to construct the relationship model between robot cute experience and user physiological indicators. Compared with the BP neural network, the relationship model established by the WOA optimized BP neural network shows lower MSE and higher R2.