Towards an intelligent problem-solving environment for signal processing

Modern signal processing methods are typically quite complex, and a great deal of expertise is needed in selecting and using appropriate methods for an application. Since most users have little signal processing experience, the expertise of the signal processing specialist needs to be made more accessible to them. This can be done by means of knowledge-based systems developed using dedicated AI tools for signal processing, as proposed in this paper. We first discuss the types of knowledge and reasoning involved in SP problem-solving, and based on this, we propose an AI architecture which facilitates the expression of knowledge at the level of the SP specialist. SP methods are represented as operators, and grouped according to the SP goals they perform. Processing is initiated by means of requests, which ask for a particular goal to be affected on a given data set. Two types of reasoning, planning and supervision, are required for the processing, and mechanisms for them are provided. The paper describes how a knowledge based system can be built using the proposed architecture, and how a given SP problem is solved by the KBS.