维特比回溯

如何解决维特比回溯

def viterbi(hmm,obs):
    # initialize empty Matrix
    V = []
    B = []
    N = len(hmm.states)
    T = len(obs)

    #initialize
    for q in range(1,N-1):
        V[q][1] = (hmm.trans_prob[0][q] * hmm.obs_likelihood[q][1])
        B[q][1] = 0
    #get probability for each state 
    for t in range(2,T):
        for q in range(1,N-1):
            V[q][t] = V[q][t-1] * hmm.trans_prob[q][q] * hmm.obs_likelihood[q][t]
            B[q][t] = V[q][t-1].argmax() * hmm.trans_prob[q][q] * hmm.obs_likelihood[q][t]

    bestpathprob = max(V[N-1][T])
    bestpathpointer = V[N-1][T].argmax()
    return bestpathpointer

如何为Viterbi函数实现回溯？回溯需要获得viterbi [（从1到N）] [T]

的最大值