This paper proposes the replacement of the ordinary output probability with its expected value if the addition of noise is modeled as a stochastic process, which in turn is merged with the hidden Markov model (HMM) in the Viterbi algorithm. This new output probability is analytically derived for the generic case of a mixture of Gaussians and can be seen as the definition of a stochastic version of the weighted Viterbi algorithm. Moreover, an analytical expression to estimate the uncertainty in noise canceling is also presented. The method is applied in combination with spectral subtraction to improve the robustness to additive noise of a text-dependent speaker verification system. Reductions as high as 30% or 40% in the error rates and improvements of 50% in the stability of the decision thresholds are reported.Index Terms-Hidden Markov model (HMM), noise robustness, speaker verification, Viterbi algorithm.

No abstract

No abstract

We report the discovery of eight new giant planets, and updated orbits for four known planets, orbiting dwarf and subgiant stars using the CORALIE, HARPS, and MIKE instruments as part of the Calan-Hertfordshire Extrasolar Planet Search. The planets have masses in the range 1.1-5.4M J 's, orbital periods from 40-2900 days, and eccentricities from 0.0-0.6. They include a double-planet system orbiting the most massive star in our sample (HD147873), two eccentric giant planets (HD128356b and HD154672b), and a rare 14 Herculis analogue (HD224538b). We highlight some population correlations from the sample of radial velocity detected planets orbiting nearby stars, including the mass function exponential distribution, confirmation of the growing body of evidence that low-mass planets tend to be found orbiting more metal-poor stars than giant planets, and a possible period-metallicity correlation for planets with masses >0.1 M J , based on a metallicity difference of 0.16 dex between the population of planets with orbital periods less than 100 days and those with orbital periods greater than 100 days.

This paper addresses the problem of temporal constraints in the Viterbi algorithm in speaker-dependent and independent tasks. The results here presented suggest that in a speaker-dependent task the introduction of temporal constraints can lead to a high improvement with additive or convolutional noise, the statistical modeling of state durations is not relevant if the max and min state duration restrictions are imposed, and truncated probability densities give better results than a metric previously proposed. Finally, word position dependent and independent temporal restrictions are compared in connected word speech recognition experiments and it is shown that the former leads to better results with the same computational load. However, duration model effect could be much less significant when the acoustic model is optimized and when the training and testing conditions are matched.