Calculates log of output probability of a dataframe for all states of a DD, an SC or a CD HMM.
More...
Calculates log of output probability of a dataframe for all states of a DD, an SC or a CD HMM.
- Author
- Jacques Duchateau
- Date
- 3/94
- Revision History:
- 3/94 - JD
- written for HMM V5.0 (ANSI-C), extracted from vit_align (V4.2).
- 5/94 - JD
- adapted for 'top_n' gaussian evaluation
- 8/94 - JD
- adapted for selection of gaussians per state; static allocations changed to dynamic ones (pvec, index)
- 9/94 - JD
- flooring of output probabilities in SC_HMM case changed: 'logminout' is now relative to the probability of the best gaussian, instead of an absolute flooring constant.
- 8/95 - KD
- flooring on the gaussian probs. instead of on the output probs.
- 8/95 - JD
- changes for multiple codebooks in continuous cases; changes for new hmm structures.
- 10/95 - JD
- added possibility to select which states to evaluate.
- 7/97 - JD
- rewritten for extended hmm structures.