There is a tiny issue in the implement of gauss algorithm. According to the paper you recommended about continuous HMM, the gauss algorithm in continous HMM.cpp should be
z _= (1.0/( sigma[i][n] * SQRT_TWO_PI )) * exp( - SQR(x[i][n]-y[j][n])/SQR(2.0_sigma[i][n]) );
Hi, Nickgillian!
There is a tiny issue in the implement of gauss algorithm. According to the paper you recommended about continuous HMM, the gauss algorithm in continous HMM.cpp should be z _= (1.0/( sigma[i][n] * SQRT_TWO_PI )) * exp( - SQR(x[i][n]-y[j][n])/SQR(2.0_sigma[i][n]) );
the "2" needs to be included in SQR too.