RUN_RECONSTRUCTION sets the parameters of the algorithms and partition the Motion Capture Data to Linear Dynamical Systems (LDSs). [A,B,C,u,x_0, cf_v ] = run_reconstruction(varargin) In this implementation, the Motion Capture Data are modeled from seven independent linear dynamical systems. More details can be found in Section 2 of [1]. The function returns the LDSs mattrices: A, B, C, the estimated input signal u that drives the LDSs, the initial condition of the LDSs x_0 and the cost function values cf_v run_reconstruction accepts the following arguments: system_order [10] Defines the order of the LDSs block_size [20] Size of the Hankel matrices used in System Identification algorithm lambda [10] weight of the l1 term in the cost function (sparsity penalty) s_const [10] energy constraint of the impulse response of the LDS internal_iter [3] number of iterations of the reweighted l1 minimization external_iter [10] number of alternations between system identification and input estimation plot_data [true] plotting intermediate results demo_sample [1] Three demo sequences are provided (Motion Capture data of Future Light Dataset Citation [1] Raptis,M. and Wnuk,K. and Stefano,S. Spike Train Driven Dynamical Models for Human Actions. In Proc. of the IEEE Conf. on Comp. Vis. and Patt. Recog. June 2010.