HMM求解观察序列概率的"前向算法"伪代码实现

//求解观察序列概率的前向算法
//A是N*N状态转移概率矩阵
//B是N*M符号发射概率矩阵
//pi是1*N是初始状态的概率分布
//T是终态时间
//O是1*T观察序列
forword_HMM(A, B，pj, T, O)
	N = A.length
	graph[2][N]//记录表格，利用滚动数组节约存储空间
	for t = 0 to (T - 1)
		for j = 0 to (N - 1)
			if t == 0
				//初始化
				graph[0][j] = pj[j] * B[j][O[0]]
			else
				sum = 0
				for i = 0 to ( N - 1)
					sum += graph[(t-1) % 2][i] * A[i][j]
				graph[t % 2][j] = sum * B[j][O[t]]
	p = 0
	for k = 0 to (N - 1)
		p += graph[(T-1)%2][k]
	return p