The role of interpretation and diagnosis in signal processing

A framework is proposed for adaptive signal processing that combines classical signal processing with adaptation criteria involving interpretation of the output signal and qualitative information about the classical signal processing itself. Interpretation of the output signal is achieved by expressing the signal at multiple levels of abstraction, capturing different levels of signal detail. A key aspect of the proposed framework is the computation of rich error descriptions by matching signal abstractions, as opposed to correlating numeric signals. Another key aspect is the encoding of qualitative information about the behavior of the classical signal processing component as processes that act on reference signal abstractions and produce output signal abstractions. A search is then conducted for the most plausible processes that explain the differences between the reference output and the actual output of the system.<>