Towards a flexible platform for voice accent and expression selection on a Healthcare Robot

Towards a flexible platform for voice accent and expression selection on a healthcare robot Aleksandar Igic1, Catherine I. Watson1, Jonathan Teutenberg2, Rie Tamagawa3, Bruce MacDonald1, Elizabeth Broadbent3 1 - Department of Electrical and Computer Engineering 2 - Department of Comuter Science 3 - Department of Psychological Medicine University of Auckland, Private Bag 92019, Auckland 1142, New Zealand aigi001@aucklanduni.ac.nz, jono@cs.auckland.ac.nz, {c.watson,e.broadbent,r.tamagawa,b.macdonald} @auckland.ac.nz Abstract In the application of robots in healthcare, where there is a requirement to communicate vocally with non-expert users, a capacity to generate intelligible and expressive speech is needed. The Festival Speech Synthesis System is used as a framework for speech generation on our healthcare robot. Ex- pression is added to speech by modifying mean pitch and pitch range parameters of a statistical model distributed with Festival. US and UK Eng- lish diphone voices are evaluated alongside a new- ly made New Zealand English accented diphone voice by human judges. Results show judges can discern different accents and correctly identify the nationality of the voice. 1. Introduction With the rapidly ageing populations in the developed world, robots are increasingly finding a use in nursing homes in assistive medical care [1][2]. In order for such robots to facilitate the needs of older and mobility restricted users from a communication point of view, more human modes of interaction need to be implemented [1]. The most natural mode of communication for humans is speech, which for a medical robot requires both speech recognition and generation capabilities. We are currently focusing on implementing a flexible robotic speech generation framework that will provide a high standard of quality and expressiveness. 2. Healthcare Robot Background We are currently developing a Healthcare robot to assist with elderly care. This multi-disciplined project involves personnel with backgrounds in Engineering, Health Psychology, Health Informatics, Nursing, and Gerontology. It involves academics and industry from both New Zealand and Korea [3] .The project is working closely with a retirement village in Auckland, where the healthcare robots are to be trialed. We have already evaluated a preliminary version of the robot with elderly users, in a blood pressure measurement task. The robot instructed users how to use a blood pressure measurement device, and reported back their measurements [4]. The robot (see Figure 1) is a mobile device with ultra sound and laser sensors for location detection. It has a screen with a talking virtual head and the face is able to convey a variety of emotions [5]. We discuss the development of the expressive face, accompanying the speech synthesis in [4]. Figure 1: Charles with blood pressure monitor. The functionality of the healthcare robot is now being extended to include location monitoring, falls detection, medication management, appointment reminders, and more vital signs measurements (pulse and blood oxygenation) [1][6]. Some other non medical uses could include, delivering weather and time information and reading the news. These roles require, in most cases, human robot interaction to take place in form of speech dialogue. The robot voice is provided by the Festival Speech Synthesis system …

Content truncated to 500 words