A Gaussian filtering for hybrid Hmms Tracks mode sequences with a bank of Kalman Filters g2 Problem: the number of possible mode sequences increases exponentially time 鸡)望 (also exponentially in components) Exponential computational complexity 2 strategies 1. pruning(truncating) 2. collapsing(merging) Hybrid Mode Estimation and Gaussian Filtering with Hybrid HMM Models 16412/6.834 Lecture,15 March2004 Pruning: Selecting relevant mode sequences Two methods K-best filtering Looking for a set of leading sequences Obtained efficiently with A* search Particle filtering Selecting trajectories probabilistically by sampling Rao-Blackwellised particle filtering Hybrid Mode Estimation and Gaussian Filtering with Hybrid HMM ModelsHybrid Mode Estimation and Gaussian Filtering with Hybrid HMM Models 16.412 / 6.834 Lecture, 15 March 2004 29 Gaussian filtering for hybrid HMMs z Tracks mode sequences with a bank of Kalman Filters Problem: the number of possible mode sequences increases exponentially in time (also exponentially in # components) Ÿ Exponential computational complexity 2 strategies: 1. pruning (truncating) 2. collapsing (merging) Hybrid Mode Estimation and Gaussian Filtering with Hybrid HMM Models 16.412 / 6.834 Lecture, 15 March 2004 30 Pruning: Selecting relevant mode sequences Two methods z K-best filtering ƒ Looking for a set of leading sequences ƒ Obtained efficiently with A* search z Particle filtering ƒ Selecting trajectories probabilistically by sampling ƒ Rao-Blackwellised particle filtering x2 x2 x2 x2 x2 x2 x2 x1 x1 x0 x2 x2 (1) (2) (3) x2 (3) (1) (2) x1 (3) (1) (4) (5) (6) (7) (8) (9)