Open Access Article

Title: From toy-like to intelligent: ARM-based cloud-injected AI companion robot

Authors: Yonggang Zhang; Xiao Tan; Xilong Qu

Addresses: Department of Science, Gansu University of Chinese Medicine, Lanzhou, 743000, China ' School of Information Technology and Management, Hunan University of Finance and Economics, Changsha, 410205, China ' School of Information Science and Engneering, Changsha Normal University, Changsha, 410100, China

Abstract: Our research in child companion robotics is evolving from simple toys to intelligent, ARM-based, cloud-connected AI companions. Historically, these robots have been limited by predefined interactions and speech recognition capabilities. Our innovation lies in the implementation of advanced transformer models, which have reduced word error rates by 12% compared to other models. Overcoming challenges posed by resource constraints, we optimised these models for ARM architecture, ensuring high performance in embedded systems. Furthermore, integrating cloud-injected AI enhances our robots' intelligence by providing access to real-time data, thus enriching conversations and adaptability. This integration, coupled with advancements in speech recognition, represents a significant leap towards interactive companions capable of engaging meaningfully with children. Our research marks the onset of a new era in child companion robotics, demonstrating the transformative potential of ARM-based, cloud-injected AI systems.

Keywords: speech processing; cloud integration; conversation relevance; transfer learning; dynamic injection.

DOI: 10.1504/IJICT.2024.139108

International Journal of Information and Communication Technology, 2024 Vol.24 No.7, pp.17 - 50

Received: 18 Jan 2024
Accepted: 14 Apr 2024

Published online: 13 Jun 2024 *