Inference of System Identification Performance from Input/Output Measurements

System identification concepts play a key role in adaptive signal processing applications like communication channel estimation, acoustic echo cancellation, active noise control, and adaptive feedback cancellation. In the development of such systems, one would often like to know what kind of performance could be expected and how it depends on the filter length before actually committing the adaptive filter design. Also, after the system has been implemented, it is sometimes necessary to monitor the performance. This paper shows that in such cases, one can infer the potential performance in several ways from input/output measurements of the primary system, taking into account the obscuring effects of additive noise.