Probabilistic expert systems for shape recognition applied to ultrasonics techniques

This paper evaluates and compares the application of different types of expert systems in object recognition by means of ultrasonic sensors. In fact, rules based expert systems and probabilistic expert systems have been evaluated in a calculation of prismatic body position and prismatic body faces recognition. This application was subsequently used to detect pieces and position of objects in a robotic system. A characterisation of the received ultrasonic signal envelope has been done by calculating several parameters to define it, without necessity of storing all its points. Afterwards, a study of the function of probability distribution for each parameter permits us to obtain the necessary information to provide sufficient knowledge for the expert system to make the distinction between the different objects of interest