Multi-channel Blind System Identification (MBSI) is a technique for estimating both an unknown input and unknown channel dynamics from simultaneous output measurements at different channels through which the input signal propagates. It is a powerful tool particularly for the identification and estimation of dynamical systems in which a sensor, for measuring the input, is difficult to place. All of the existing MBSI algorithms, however, are not applicable to multi-channel systems sharing common dynamics among the channels, since these algorithms, by nature, exploit "differences" among the multiple channel dynamics. This requirement renders the MBSI algorithms useless in systems that have both a lumped-parameter nature and a distributed nature; all channels in a system of this type share poles dictated by the lumped-parameter dynamics. To overcome this difficulty, this thesis investigates a new approach, Intermediate Input Identification (IIID). This thesis proves that the distinct dynamics in each channel can be identified up to a scalar factor even when common dynamics are present. Based on this discovery, the MBSI problem is reformulated and an intermediate input is introduced, which integrates the original system input and the common dynamics shared by all the channels. The two-step IIID approach is developed to solve the problem: first, the distinct dynamics are identified from the outputs; second, the common dynamics are identified from the intermediate input by exploiting the zero-input response of the system. The identifiability conditions are thoroughly investigated. The sufficient and necessary conditions and the relationship between the linear-complexity condition of the original input and that of the intermediate input are derived in this thesis.
展开▼
