FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W767081029> ?p ?o ?g. }

• W767081029 abstract "A standard way of improving the robustness of speech recognition systems to noise is model compensation. is replaces a speech recogniser’s distributions over clean speech by ones over noise-corrupted speech. For each clean speech component,model compensation techniques usually approximate the corrupted speech distribution with a diagonal-covariance Gaussian distribution. is thesis looks into improving on this approximation in two ways: rstly, by estimating full-covariance Gaussian distributions; secondly, by approximating corrupted-speech likelihoods without any parameterised distribution. e rst part of this work is about compensating for within-component feature correlations under noise. For this, the covariancematrices of the computed Gaussians should be full instead of diagonal. e estimation of oš-diagonal covariance elements turns out to be sensitive to approximations. A popular approximation is the one that state-of-the-art compensation schemes, likevts compensation, use for dynamic coe›cients: the continuous-time approximation. Standard speech recognisers contain both per-time slice, static, coe›cients, and dynamic coe›cients, which represent signal changes over time, and are normally computed from a window of static coe›cients. To remove the need for the continuous-time approximation, this thesis introduces a new technique. It rst compensates a distribution over the window of statics, and then applies the same linear projection that extracts dynamic coe›cients. It introduces a number ofmethods that address the correlation changes that occur in noisewithin this framework. e next problem is decoding speed with full covariances. is thesis reanalyses the previously-introduced predictive linear transformations, and shows how they can model feature correlations at low and tunable computational cost. e second part of this work removes the Gaussian assumption completely. It introduces a sampling method that, given speech and noise distributions and a mismatch function, in the limit calculates the corrupted speech likelihood exactly. For this, it transforms the integral in the likelihood expression, and then applies sequential importance resampling. ough it is too slow to use for recognition, it enables a more ne-grained assessment of compensation techniques, based on the kl divergence to the ideal compensation for one component. e kl divergence proves to predict the word error rate well. is technique also makes it possible to evaluate the impact of approximations that standard compensation schemes make." @default.
• W767081029 created "2016-06-24" @default.
• W767081029 creator A5034178766 @default.
• W767081029 creator A5040952492 @default.
• W767081029 date "2011-11-08" @default.
• W767081029 modified "2023-09-23" @default.
• W767081029 title "Statistical models for noise-robust speech recognition" @default.
• W767081029 cites W105750483 @default.
• W767081029 cites W124847279 @default.
• W767081029 cites W127709214 @default.
• W767081029 cites W141296123 @default.
• W767081029 cites W1486632395 @default.
• W767081029 cites W1502469360 @default.
• W767081029 cites W1530235965 @default.
• W767081029 cites W1543510152 @default.
• W767081029 cites W1550073879 @default.
• W767081029 cites W1553154706 @default.
• W767081029 cites W1561429093 @default.
• W767081029 cites W1573570773 @default.
• W767081029 cites W1585285631 @default.
• W767081029 cites W158643468 @default.
• W767081029 cites W1599512239 @default.
• W767081029 cites W1643320849 @default.
• W767081029 cites W1663973292 @default.
• W767081029 cites W1665196592 @default.
• W767081029 cites W167557053 @default.
• W767081029 cites W169377849 @default.
• W767081029 cites W174003511 @default.
• W767081029 cites W1864449204 @default.
• W767081029 cites W1898362271 @default.
• W767081029 cites W191461873 @default.
• W767081029 cites W1966737826 @default.
• W767081029 cites W1974387177 @default.
• W767081029 cites W1976587308 @default.
• W767081029 cites W1991133427 @default.
• W767081029 cites W1994396704 @default.
• W767081029 cites W2002342963 @default.
• W767081029 cites W2003486321 @default.
• W767081029 cites W2010023285 @default.
• W767081029 cites W2018519656 @default.
• W767081029 cites W2020999234 @default.
• W767081029 cites W2033178790 @default.
• W767081029 cites W2035328943 @default.
• W767081029 cites W2049633694 @default.
• W767081029 cites W2057349863 @default.
• W767081029 cites W2063291689 @default.
• W767081029 cites W2067117291 @default.
• W767081029 cites W2068772561 @default.
• W767081029 cites W2069631319 @default.
• W767081029 cites W2077611006 @default.
• W767081029 cites W2086699924 @default.
• W767081029 cites W2101686922 @default.
• W767081029 cites W2104448323 @default.
• W767081029 cites W2106554350 @default.
• W767081029 cites W2106714211 @default.
• W767081029 cites W2109349638 @default.
• W767081029 cites W2110575115 @default.
• W767081029 cites W2111093880 @default.
• W767081029 cites W2111421849 @default.
• W767081029 cites W2112089769 @default.
• W767081029 cites W2114325887 @default.
• W767081029 cites W2118497033 @default.
• W767081029 cites W2118984443 @default.
• W767081029 cites W2120025343 @default.
• W767081029 cites W2121981798 @default.
• W767081029 cites W2122009793 @default.
• W767081029 cites W2122699724 @default.
• W767081029 cites W2123487311 @default.
• W767081029 cites W2126415164 @default.
• W767081029 cites W2128653836 @default.
• W767081029 cites W2135346934 @default.
• W767081029 cites W2136160685 @default.
• W767081029 cites W2140595311 @default.
• W767081029 cites W2140604339 @default.
• W767081029 cites W2142980177 @default.
• W767081029 cites W2146871184 @default.
• W767081029 cites W2148154194 @default.
• W767081029 cites W2150660184 @default.
• W767081029 cites W2151484683 @default.
• W767081029 cites W2155445312 @default.
• W767081029 cites W2156450495 @default.
• W767081029 cites W2157590573 @default.
• W767081029 cites W2158195707 @default.
• W767081029 cites W2159283619 @default.
• W767081029 cites W2163680580 @default.
• W767081029 cites W2165053039 @default.
• W767081029 cites W2167270514 @default.
• W767081029 cites W2170246298 @default.
• W767081029 cites W2187805118 @default.
• W767081029 cites W2280467666 @default.
• W767081029 cites W2395140404 @default.
• W767081029 cites W2411442511 @default.
• W767081029 cites W289901950 @default.
• W767081029 cites W29946472 @default.
• W767081029 cites W31925794 @default.
• W767081029 cites W37220840 @default.
• W767081029 cites W74241614 @default.
• W767081029 cites W75761934 @default.
• W767081029 cites W1548718523 @default.
• W767081029 doi "https://doi.org/10.17863/cam.14011" @default.

• next