Development of a NAO Humanoid based Medical Assistant

Abstract

In this paper1, socially assistive human-robot interaction has been explored on a NAO Humanoid Robot in order to automate the pharmacy and bio-medical sector, with a broader aim of addressing all similar tasks. The problem has been divided into three sub-segments viz. pick and place operation with smooth gripping mechanism, reading printed and handwritten text from prescriptions, and use of smart detection technology, with a focus on barcode detection to locate target objects (medicine flaps) in real time. Iterative Jacobian Pseudo inverse kinematics algorithm is implemented to calculate the joint angles. To account for the poor performance of Google Tesseract for handwritten text, the image contrast is enhanced for histogram equalization and fed to maximally stable extremal regions (MSER) algorithm in a combination with Stroke Width Transform (SWT) to make text detection more robust even in presence of blur, before feeding it to Tesseract. Lastly, two techniques are developed to incorporate barcode integration with NAO, first using ALBarcodeReader API, the limitations of which are solved using vertical descent sobel scharr operator to attain real time barcode scanning for all types of barcodes.