Empathetic Robot for the Elderly using Machine Learning

Abstract

According to WHO estimates, there is a growing population of over 1 billion people aged above 60 years of age. There has been an increasing shortage of caregivers for the aging populations. This has opened a market of USD 7 billion dollars in senior care alone. In a struggle to care for our elderly and capitalize from this industry in efficiency, this paper presents an empathetic robot that can make day-to-day activities hassle free. It is a voice assistant that can detect the emotion of the speaker and reply with contextual awareness to bridge the superstitious gap between the elderly and technology. It achieves this with multi-modal classification of emotion using audio and text. i) Machine learning model using Librosa to engineer features and Support vector classifier (SVC) , ii) BERT based model using transfer learning to categorize text. The robot can also set reminders for appointments and medicine intake to help with forgetfulness. The robot is customizable by a caregiver or loved one through the web application where details of any messages, reminders, descriptions of prescriptions could be entered. The robot comes with a pill dispenser that can rotate and dispense pills at the correct time and also notify the senior through the speaker. A raspberry pi is used to convert the speech to text and vice versa. The design of this robot paves the path to providing a realistic, care-giving experience to the elderly.