hmm.h

本程序中列举了数字图象模式识别领域中的常用程序,可以在其基础上进行图象处理.

//hmm.h --hidden markov model

#ifndef _hmm_h
#define _hmm_h

#include <stdlib.h>
#include <stdio.h>
#include <math.h>

typedef struct {
	int N;		/* number of states;  Q={1,2,...,N} */
	int M; 		/* number of observation symbols; V={1,2,...,M}*/
	double	**A;	/* A[1..N][1..N]. a[i][j] is the transition prob
			   of going from state i at time t to state j
			   at time t+1 */
	double	**B;	/* B[1..N][1..M]. b[j][k] is the probability of
			   of observing symbol k in state j */
	double	*pi;	/* pi[1..N] pi[i] is the initial state distribution. */
} HMM;
void ReadHMM(FILE *fp, HMM *phmm);
void PrintHMM(FILE *fp, HMM *phmm);
void InitHMM(HMM *phmm, int N, int M, int seed);
void CopyHMM(HMM *phmm1, HMM *phmm2);
void FreeHMM(HMM *phmm);

void ReadSequence(FILE *fp, int *pT, int **pO);
void PrintSequence(FILE *fp, int T, int *O);
void GenSequenceArray(HMM *phmm, int seed, int T, int *O, int *q);
int GenInitalState(HMM *phmm);
int GenNextState(HMM *phmm, int q_t);
int GenSymbol(HMM *phmm, int q_t);

 
void Forward(HMM *phmm, int T, int *O, double **alpha, double *pprob);
void ForwardWithScale(HMM *phmm, int T, int *O, double **alpha,
        double *scale, double *pprob);
void Backward(HMM *phmm, int T, int *O, double **beta, double *pprob);
void BackwardWithScale(HMM *phmm, int T, int *O, double **beta,
        double *scale, double *pprob);

//////////////////////////////////////////////////////////////////////////////
// 函数：估计HMM参数，保存在phmm中
// 输入参数为特征序列O，长度为T
// 一般在调用前需要做以下alloc memory的工作，下为示例： 
//alpha = dmatrix(1, T, 1, hmm.N)
//	beta = dmatrix(1, T, 1, hmm.N)
//	gamma = dmatrix(1, T, 1, hmm.N)
//	BaumWelch(&hmm, T, O, alpha, beta, gamma, &niter, 
//		&logprobinit, &logprobfinal)
// 结束后释放内存，示例如下：
// free_dmatrix(alpha, 1, T, 1, hmm.N)
//free_dmatrix(beta, 1, T, 1, hmm.N)
//	free_dmatrix(gamma, 1, T, 1, hmm.N)
void BaumWelch(HMM *phmm, int T, int *O, double **alpha, double **beta,
        double **gamma, int *niter, 
	double *plogprobinit, double *plogprobfinal);

double *** AllocXi(int T, int N);
void FreeXi(double *** xi, int T, int N);
void ComputeGamma(HMM *phmm, int T, double **alpha, double **beta,
        double **gamma);
void ComputeXi(HMM* phmm, int T, int *O, double **alpha, double **beta,
        double ***xi);

////////////////////////////////////////////////////////////////////////
// 下面两个函数都是利用Viterbi方法估计在phmm模型参数下，特征序列
// 为O的概率，其中T为序列长度，结果保存在pprob中
// 一般在调用前申请如下内存
//	q = ivector(1,T);
//delta = dmatrix(1, T, 1, hmm.N)
//psi = imatrix(1, T, 1, hmm.N)
//当然了，这些内存还是要释放的，q是相应的状态变化序列，长度也是T
void Viterbi(HMM *phmm, int T, int *O, double **delta, int **psi,
        int *q, double *pprob);
void ViterbiLog(HMM *phmm, int T, int *O, double **delta, int **psi,
        int *q, double *pprob);

/* random number generator related functions*/

int hmmgetseed(void);
void hmmsetseed(int seed);
double hmmgetrand(void);
 
#ifndef MAX
#define MAX(x,y)        ((x) > (y) ? (x) : (y))
#endif

#ifndef MIN
#define MIN(x,y)        ((x) < (y) ? (x) : (y))
#endif

#endif