The work aims to explore the effects of interaction level, pitch and gender on perceived anthropomorphism, usefulness, and satisfaction of the outpatient service assistant to guide the image and speech design of voice assistants in outpatient service scenarios and more public places. Twelve participants performed conversation tasks with voice assistants of different interaction level, pitch and gender in simulated outpatient service scenarios. Experiments were carried out by Wizard of Oz technique. A seven-point Likert scale was used to obtain ratings of perceived anthropomorphism, usefulness, satisfaction, and interviews were conducted. The results were analyzed by Friedman test, Wilcoxon signed rank test, and least square regression to explore users' preferences on image and speech of outpatient service assistants. Finally, according to the research results, the design of a multi-modal user interface of the outpatient service system was carried out. The results showed that in the outpatient service scenarios, users preferred outpatient service assistants with high interaction level. In terms of pitch, low-pitched female voice and high-pitched male voice were rated higher. Yet, there was no significant difference in the user's gender preferences. Satisfaction increased with the increase of perceived anthropomorphism. For voice interaction design in public places, all kinds of users should be considered to design the voice assistant that can be accepted by most users, thus taking advantage of the multi-modal user interface.