HMM-embedded Bayesian network for heterogeneous command integration: applications to biped humanoid operation over the network

Abstract

A method to combine a Bayesian Network (BN) and Hidden Markov Models (HMMs) is presented. This compound system is applied to robot operations. The addressed problem and presented methods are novel with the following features: (1) BN and HMMs make a total decision system by accepting evidences from HMMs to the BN. (2) The HMM-embedded BN is applied to the human motion recognition for the biped humanoid operation. (3) Besides the motion recognition, the image recognition is incorporated by adding a BN subsystem. Thus, the total HMM-embedded BN can be regarded as an integrator of heterogeneous commands. (4) The human operator and the biped humanoid can be located on the other side of the network each other. (5) The piped humanoid follows various commands of human motions without falling down by showing better sophistication and operation success than HMM-alone and BN-alone systems. In addition to the above, an information supply to the BN from brain signals is realized through a combination with a Support Vector Machine (SVM).