ASR and TTS for Voice Controlled Child-Robot Interactions in Italian: Empirical Study Findings on the Aliz-e Project ...

Abstract. Artificial companion agents are becoming increasingly important in the field of healthcare, particularly when children are involved, with the aim of providing novel educational tools, supporting communication between young patients and hospital personnel and taking on the role of entertainment robots. The principal application of the European FP7 project Aliz-e is the development of an educational robot companion for children (target age 8-11) who are affected by metabolic disorders such as diabetes and/or obesity. The purpose of this educational robot is to enhance the child's well-being and facilitate therapeutic activities in a hospital setting. Though speech comprehension, in particular acoustic analysis applied to automatic speech recognition of children's voices, has been investigated extensively by speech technology researchers over the last two decades, most of the literature is focused on the English language. Given that the primary evaluation site of the Aliz-e project is located in an Italian hospital, ISTC-CNR researchers have focused on the application of speech technologies in the Italian language as spoken by children. This chapter also outlines the investigation of voice adaptation techniques in children ASR. It reports on an experiment of ASR in a real case child-robot interaction scenario in a hospital setting, and presents the data collection for a corpus of annotated Italian children's speech. The study also shows how in the production of speech the robotic companion must be able to convey to the child the identity and the emotional state of the speaker, in addition to verbal content. In addition, the robot companion must focus on particular words that are important in the communication with the child. This chapter describes the tools and the modules needed to build a Text to Speech engine implementing these features designed for the Italian language.