A proposal of an internal-state inference system based on multimodal sensory fusion method

A system that infers the internal state of target objects such as cans based on the objects´ multi-modal response against a given stroke and fuzzy inference is proposed. A proposed system has a hammer by which the target object is struck and its response against the stroke is observed with both a microphone for auditory information and an accelerometer attached to the hammer. The internal state inference algorithm is based on the fuzzy matching using fuzzy membership data stored in advance. Experimental results show the proposed system can discriminate a plastic can from a metal can with 100% accuracy, while three types of internal state (amount of water empty, half, full) can be recognized with 80 % accuracy. This paper gives the foundation of low-cost and nondestructive automatic diagnosis systems