TOYBOT: An Interactive exercise companion for rehabilitation of sedentary geriatric and obese children

Abstract

Differently abled children usually have difficulty following the vocal instructions of a physiotherapist while undergoing exercise rehabilitation sessions. This makes the job of therapists tiresome and creates delay in the progressive development of the child. In addition, the geriatric category and obese children often find themselves lonely or less motivated to start and continue with a physically active lifestyle. Both these group of people require a device, a Toybot, which act as a companion and encourage them to do a set of pre-defined exercises, which are demonstrated by the device. The paper discusses the design and evaluation of a Socially Assitive Robotic (SAR) rehabilitation toy which can demonstrate a set of aerobic exercises and provides corrective feedback according to the participant's performance as monitored by an RGB camera. The system architecture consists of three modules: Vision, Voice, and Motion modules. Vision module recognizes the presence of the user by face detection, adjusts the position in the line of sight, and starts capturing the posture of the user while following the exercises demonstrated by the Toybot. The real-time posture detection is implemented with a ML Kit Pose Detection API. The speech module is in charge of converting text strings into spoken words, to give greetings, feedback, and corrective instructions, implemented with the help of VoiceText TTS Engine on a Windows 10 based laptop. The action module is in charge of control of HS-65MG servomotors to demonstrate various exercises such as aerobic flexion, rotation and abduction of the upper limbs and neck. All three modules are under the coordination of an ATMEGA328P based Arduino microcontroller. User feedback during the evaluation of the toy suggests that social interaction between the robot and the user is essential and advantageous for sustaining user engagement and influencing intrinsic motivation.