A Flexible Framework for Expectation Maximization-Based MIMO System Identification for Time-Variant Linear Acoustic Systems

Quasi-continuous system identification of time-variant linear acoustic systems can be applied in various audio signal processing applications when numerous acoustic transfer functions must be measured. A prominent application is measuring head-related transfer functions. We treat the underlying multiple-input-multiple-output (MIMO) system identification problem in a state-space model as a joint estimation problem for states, representing impulse responses, and state-space model parameters using the expectation maximization (EM) algorithm. We address limitations of prior work by imposing different model structures, especially for dependencies within a (transformed) state vector. This results in block diagonal matrix structures, for which we derive M-step update rules. Making assumptions about this model structure and choosing a block size for a given application define the computational complexity. In examples, we found that applying this framework yields improvements of up to 10 dB in relative system distance in comparison to a conventional method.