Professional Documents
Culture Documents
Score
Score
Score
SCORING PROCEDURE For each speech unit the log-likelihood ratio (LLR) that this unit has been uttered by the claimant speaker is computed. Because LLR's for utterances yielded virtually 100% correct identication, we decided to compute LLR's for subwords and words too. LLR's for words and utterances were obtained by summing the LLR's for the subwords making up the larger units.
From page score1 A technique for improved score calculation and normalization in a framework of recognition with phonetically structured speaker models. The technique involves determining, for each frame and each level of phonetic detail of a target speaker model, a non-interpolated likelihood value, and then resolving the at least one likelihood value to obtain a likelihood score.
A score is calculated from a sequence of observations y 1 , . . . , y n extracted from the speech input. The scoring function is based on the output of a generalized linear discriminant function of the form g(y) = w t b(y), where w is the vector of classier parameters (model) and b is an expansion of the input space into a vector of scalar functions.
and CilliF.pdf
speakers. This is done by computing a match score. This score is a measure of the similarity between the input feature vectors and some model. There are two types of models: template models and stochastic models. In template models, the pattern matching is deterministic. The align ment of the observed frames to template frames is selected to minimize a distance measure value. In stochastic models, the pattern matching is probabilistic. The result is a measure of the likelihood, or conditional probability, of the observation given the model. [5]. DTW is one of the most important template modeling techniques and HMM is one of the most important stochastic modeling techniques.