000220896 001__ 220896
000220896 005__ 20190317000514.0
000220896 037__ $$aCONF
000220896 245__ $$aEfficient Posterior Exemplar Search Space Hashing Exploiting Class-Specific Sparsity Structures
000220896 269__ $$a2016
000220896 260__ $$c2016
000220896 336__ $$aConference Papers
000220896 520__ $$aThis paper shows that exemplar-based speech processing using class-conditional posterior probabilities admits a highly effective search strategy relying on posteriors' intrinsic sparsity structures. The posterior probabilities are estimated for phonetic and phonological classes using deep neural network (DNN) computational framework. Exploiting the class-specific sparsity leads to a simple quantized posterior hashing procedure to reduce the search space of posterior exemplars. To that end, small subset of quantized posteriors are regarded as representatives of the posterior space and used as hash keys to index subsets of similar exemplars. The $k$ nearest neighbor ($k$NN) method is applied for posterior based classification problems. The phonetic posterior probabilities are used as exemplars for phoneme classification whereas the phonological posteriors are used as exemplars for automatic prosodic event detection. Experimental results demonstrate that posterior hashing improves the efficiency of $k$NN classification drastically. This work encourages the use of posteriors as discriminative exemplars appropriate for large scale speech classification tasks.
000220896 6531_ $$aAutomatic prosodic event detection
000220896 6531_ $$aFast $k$NN
000220896 6531_ $$aPhoneme classification
000220896 6531_ $$aPosterior representatives
000220896 6531_ $$aQuantized posterior hashing
000220896 6531_ $$aStructured sparsity
000220896 700__ $$0243353$$g188259$$aAsaei, Afsaneh
000220896 700__ $$aLuyet, Gil
000220896 700__ $$aCernak, Milos
000220896 700__ $$aBourlard, Hervé
000220896 7112_ $$cSan Francisco, CA$$aInterspeech
000220896 8564_ $$uhttp://publications.idiap.ch/index.php/publications/showcite/Asaei_Idiap-RR-10-2016$$zRelated documents
000220896 8564_ $$uhttps://infoscience.epfl.ch/record/220896/files/Asaei_INTERSPEECH_2016.pdf$$zn/a$$s304766$$yn/a
000220896 909C0 $$xU10381$$0252189$$pLIDIAP
000220896 909CO $$ooai:infoscience.tind.io:220896$$qGLOBAL_SET$$pconf$$pSTI
000220896 937__ $$aEPFL-CONF-220896
000220896 970__ $$aAsaei_INTERSPEECH_2016/LIDIAP
000220896 973__ $$aEPFL
000220896 980__ $$aCONF