Multi-Modal Humanoid Robot

Abstract

Human-Robot Interaction deals with the direct use of robotic systems to interact with humans in particular context. In this paper, a humanoid is developed which can understand the commands in the form of speech and gesture. A connected-word speaker-independent speech recognition system is built using Mel Frequency Cepstral Coefficient and Gaussian Mixture Model in Kaldi toolkit. Gesture recognition is implemented using Convolutional Neural Network.