Evaluating the Voice Interface of Conversational Speech Systems in Human-Robot Interaction for Elderly People

Abstract

Conversational speech systems have been actively studied in recent years. However, their role as a conversational agent in daily living is still small. Voice interaction technology, which is supposed to play many roles, is not actively used. Voice interaction requires further research and development in terms of Human-Robot Interaction regarding its actual use. <br/> Conversational agents are stand-alone devices that do not connect to other devices, unlike existing smart devices. Voice interaction would improve the access of the elderly because it is difficult for elderly people to interact using methods such as a computer mouse, keyboards, and touch screens. Voice-based interactions are the most usual method by which people express their thoughts and needs. In addition, the voice command input methods are intelligent and natural, and no additional input devices are needed. Based on these advantages, it is expected that conversational agents will be useful for the elderly in the near future. These technologies will be useful for enhancing the independence of the elderly and their quality of life as much as possible. Conversational speech systems are an assistive technology that has valuable possibilities in helping the elderly sustain their independence and improve their welfare. <br/> Despite the above advantages, the user experience of conversational speech systems has not been fully considered. Hence, research on how voice interaction technology will be accepted by users is of paramount importance. The user experience of products or services for the elderly should be thoroughly investigated to improve the elderlys experience of use. This study aimed to evaluate the user experience of a conversational speech system and suggest voice interaction design guidelines. <br/> In the experiment, 53 subjects (25 male and 28 female) participated. Their ages ranged from 60 to 82 years (mean = 69.14). The experimental laboratory was divided by a partition on opposite sides of which the experimenter and subjects were placed. On the subjects side was a speaker that functioned as a conversational speech system, while there was a computer on the experimenters side to send the recorded voice files to the speaker. This study designed the Wizard of Oz method for this experiment so that the subjects would believe that the automatic conversational speech system allows spontaneous conversations to occur, although in fact the experimenter was controlling the conversation system. <br/> This study investigated the user characteristics of the elderly and the task characteristics of the conversational speech system. Regression analysis was used to examine how the UX factor affects the satisfaction of different users differently. The experimental results show that all the major influential UX factors in each user group and task type were different. Therefore, it was possible to identify which UX factors voice interaction should be focused on according to the characteristics of elderly people and task characteristics of the system. Additionally, experimental video images were analyzed and voice interaction factors were extracted and measured. The voice interaction factors were correlated with the UX factors, which made it possible to determine the interaction factors that significantly correlated with the UX factors. The investigation of the voice interaction factor is a guide that allows the UX factors to be used directly in the interaction design. <br/> This study proposed a design model of a conversational speech system that suggests directions for designing a conversational system. Finally, eight design models emerged: six for user and two for task characteristics. The results were verified in actual contexts. An empirical evaluation of the design model was conducted, and an objective evaluation of various aspects of the results was performed using new data. The design guidelines summarize the results of this study and suggest optimized design values for each context. <br/> Through the development of a conversational speech system based on these results, it will be possible to optimize elderly peoples experience of using the system. The interaction design reflecting this guideline will maximize the elderly individuals user experience and satisfaction with the use of the conversational speech system.