HMM Basic Problem 2 Viterbi Decoding Same principle as forward algorithm with an extra term iNitialization tErmination 6(i)=x;p(z1|x;) P=max[5(l v1(i)=0 rg 2)Induction 4) Back tracking 2()RNOp(x1x,a)小v=x) 1+1(x v, =argmax8, ()p(x, |x, a)] Isis What you should know Form of Bayes Filter Kalman filter representation How to take a state estimation problem and represent it as a Kalman(or Extended Kalman) filter What terms are needed and how to find them What a hidden Markov model is The Forward algorithm The Viterbi algorithm3) Termination 4) Back tracking HMM Basic Problem 2 Ɣ algorithm, with an extra term 1) Initialization 2) Induction ( ) 0 ( ) ( | ) 1 1 1 i i x i i \ G S > @ ( ) > @ ( ) ( | , ) ( ) max ( ) ( | , ) ( | ) 1 1 | | 1 1 | | t i j j j X t t i j j t i j X t i j x a i j x a x   \ G G G > @ > @ ( ) max ( ) 1 | | * 1 | | x i P i T i X T T i X G G ( ) * 1 1 * t t t x \ x What you should know Ɣ Form of Bayes’ Filter Ɣ Ɣ How to take a state estimation problem and represent Ɣ What terms are needed and how to find them Ɣ What a Hidden Markov Model is Ɣ The Forward algorithm Ɣ Viterbi Decoding: Same principle as forward p z arg max p x p x p z d d d d arg max d d d d Kalman filter representation it as a Kalman (or Extended Kalman) filter The Viterbi algorithm