Identifying IIR filter coefficients using particle swarm optimization with application to reconstruction of missing cardiovascular signals

• We introduce a methodology to reconstruct missing segments of multidimensional data.
• IIR transfer function coefficients are identified using particle swarm optimization.
• The method was successfully applied to intensive care unit (ICU) dataset.
• Advantages are simplicity, fast computation and good results on available datasets.
• The method is trivial to parallelize and it would be straight forward to apply online.

In this paper, a new methodology to reconstruct missing segments of multidimensional data is introduced. Using the signals׳ history we identify the interconnections between the signals in the form of composite infinite impulse response (IIR) transfer functions. Assuming that the connections do not vary in time, we managed to reconstruct the missing signals using the yet available parallel measured signals and the transfer functions. The filter coefficients are fitted to the data using particle swarm optimization (PSO). The method was successfully applied to intensive care unit (ICU) data. Besides simplicity, short running time and good results on available datasets, further advantages of the proposed method are trivial parallelization and the possibility of application in an online setting.