Filters








2,189 Hits in 3.8 sec

Context-Dependent Multiple Distribution Phonetic Modeling with MLPs

Michael Cohen, Horacio Franco, Nelson Morgan, David E. Rumelhart, Victor Abrash
1992 Neural Information Processing Systems  
In this paper. we present a new MLP architecture and training algorithm which allows the modeling of context-dependent phonetic classes in a hybrid MLP/HMM: framework.  ...  The new training procedure smooths MLPs trained at different degrees of context dependence in order to obtain a robust estimate of the cootext-dependent probabilities.  ...  Acknowledgements The work reported here was partially supported by DARPA Contract MDA904-9O-C-5253. Discussions with Herve Bourlard were very helpful.  ... 
dblp:conf/nips/CohenFMRA92 fatcat:xlhcer7gwzawfdom5ak62pcpzq

NeuralNetwork-Viterbi: A Framework for Weakly Supervised Video Learning [article]

Alexander Richard, Hilde Kuehne, Ahsan Iqbal, Juergen Gall
2018 arXiv   pre-print
On several action segmentation benchmarks, we obtain an improvement of up to 10% compared to current state-of-the-art methods.  ...  Since even a small amount of videos easily comprises several million frames, methods that do not rely on a frame-level annotation are of special importance.  ...  Moreover, we showed that using an explicit length model and optimizing the video classes directly leads to a huge improvement over related HMM-based methods that use a pseudo ground-truth.  ... 
arXiv:1805.06875v1 fatcat:juipg3pm25hhdd6xurgopzjoji

NeuralNetwork-Viterbi: A Framework for Weakly Supervised Video Learning

Alexander Richard, Hilde Kuehne, Ahsan Iqbal, Juergen Gall
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
On several action segmentation benchmarks, we obtain an improvement of up to 10% compared to current state-of-the-art methods.  ...  Since even a small amount of videos easily comprises several million frames, methods that do not rely on a frame-level annotation are of special importance.  ...  Moreover, we showed that using an explicit length model and optimizing the video classes directly leads to a huge improvement over related HMM-based methods that use a pseudo ground-truth.  ... 
doi:10.1109/cvpr.2018.00771 dblp:conf/cvpr/RichardKIG18 fatcat:ufj6n4qqs5a4zddnx464ebzyfm

Writer independent on-line handwriting recognition using an HMM approach

Jianying Hu, Sok Gek Lim, Michael K. Brown
2000 Pattern Recognition  
A combination of signal normalization preprocessing and the use of invariant features makes the system robust with respect to variability among di!erent writers as well as di!  ...  A combination of point oriented and stroke oriented features yields improved accuracy. Language modeling constrains the hypothesis space to manageable levels in most cases.  ...  Therefore, including invariant features can greatly improve the robustness of the system.  ... 
doi:10.1016/s0031-3203(99)00043-6 fatcat:ntrp33tn4bfsvelt7y54m74yty

Survey on classifying human actions through visual sensors

Michael S. Del Rose, Christian C. Wagner
2011 Artificial Intelligence Review  
are working towards this by identifying new algorithms for solving parts of this problem.  ...  The ability to predict the intentions of people based solely on their visual actions is a skill only performed by humans and animals.  ...  In this case, phrase grammars limit the search set of words to improve the accuracy of what is being described. They also speed up the process over not using grammars.  ... 
doi:10.1007/s10462-011-9232-z fatcat:wp737zlzknhutjqcwm3ip3mfpu

LeRec: A NN/HMM Hybrid for On-Line Handwriting Recognition

Yoshua Bengio, Yann LeCun, Craig Nohl, Chris Burges
1995 Neural Computation  
The preprocessor performs a word-level normalization by tting a model of the word structure using the EM algorithm.  ...  The recognizer is a convolution network which can be spatially replicated. From the network output, a hidden Markov model produces word scores.  ...  Y.B. would also like acknowledge the support of the National Research and Engineering Research Council of Canada.  ... 
doi:10.1162/neco.1995.7.6.1289 pmid:7584903 fatcat:tlwq2o7g2jfhjnaagp5w7ntx3u

Combining high-level features with sequential local features for on-line handwriting recognition [chapter]

Jianying Hu, Amy S. Rosenthal, Michael K. Brown
1997 Lecture Notes in Computer Science  
This method allows incorporation of information carried by high-level long-range features while at the same time maintains the high reliability of the recognition system.  ...  We report experimental results on an HMM based recognizer for writer independent recognition of unconstrained handwritten words.  ...  Each arc in the grammar network corresponds to a letter model representing a unique letter pattern class.  ... 
doi:10.1007/3-540-63508-4_179 fatcat:drkodvtrmrfkjmzfvkzrcjo7zi

Biologically-Inspired Spike-Based Automatic Speech Recognition of Isolated Digits Over a Reproducing Kernel Hilbert Space

Kan Li, José C. Príncipe
2018 Frontiers in Neuroscience  
Compared to HMM using Mel-frequency cepstral coefficient (MFCC) front-end without time-derivatives, our MFCC-KAARMA offered improved performance.  ...  Moreover, we show that this novel framework can outperform both traditional hidden Markov model (HMM) speech processing as well as neuromorphic implementations based on spiking neural network (SNN), yielding  ...  We are also thankful to the editor and reviewers for their valuable comments and suggestions that improved the manuscript.  ... 
doi:10.3389/fnins.2018.00194 pmid:29666568 pmcid:PMC5891646 fatcat:weji6gclmzbrjl5frmbipfmjwy

Using Language Modelling to Integrate Speech Recognition with a Flat Semantic Analysis

Dirk Bühler, Wolfgang Minker, Artha Elciyanti
2005 SIGDIAL Conferences  
In this paper, we present a simple one-stage decoding scheme that can be realised without the implementation of a specialized decoder, nor the use of complex language models.  ...  Instead, we reduce an HMMbased semantic analysis to the problem of deriving annotated versions of the conventional language model, while the acoustic model remains unchanged.  ...  Usually a back-off and discounting strategy is applied in order to improve robustness in the face of unseen events.  ... 
dblp:conf/sigdial/BulerME05 fatcat:fu3ujobpnzexhklw6mbwi2zlka

Connectionist probability estimators in HMM speech recognition

S. Renals, N. Morgan, H. Bourlard, M. Cohen, H. Franco
1994 IEEE Transactions on Speech and Audio Processing  
In conclusion, we show that a connectionist component improves a state-of-the-art HMM system.  ...  We review the basis of HMM speech recognition and point out the possible benefits of incorporating connectionist networks.  ...  Robinson, and an anonymous referee for a critical reading of the manuscript.  ... 
doi:10.1109/89.260359 fatcat:y6y7wsss6fcotp4bei7x3qf2ju

Continuous speech recognition using hidden Markov models

J. Picone
1990 IEEE ASSP Magazine  
In this paper, w e review the use of Markov models in continuous speech recognition. Markov models are presented as a generalization of i t s predecessor technology, Dynamic Programming.  ...  Markov modeling provides a mathematically rigorous approach to developing robust statistical signal models.  ...  Competing hypotheses can be tracked using stack decoding [51] , by examining error patterns for the training database [521, by building grammars that explicitly generate the confusion class [41] , or  ... 
doi:10.1109/53.54527 fatcat:4qr2kggggvgflaq3ti3kxbbq3y

Hidden Markov model-based supertagging in a user-initiative dialogue system

Jens Bäcker, Karin Harbusch
2002 International Workshop on Tree Adjoining Grammars and Related Formalisms  
The entire design of a Hidden Markov-based Supertagger with trigrams builds the central issue of this paper. The evaluation of our German Supertagger lags behind the English one.  ...  Some of the reasons will be addressed later on.  ...  First, a hierarchy of neural networks classifies the user's whole turn ( on average 25 words) according to a list of problem classes.  ... 
dblp:conf/tag/BackerH02 fatcat:k7eaweqsivamzdpai26hmzun6e

Rapid development of spoken language understanding grammars

Ye-Yi Wang, Alex Acero
2006 Speech Communication  
While the focus of SGStudio is to increase productivity, experimental results show that it also improves the quality of the grammars being developed.  ...  We focus on the underlying technology of SGStudio, including knowledge assisted example-based grammar learning, grammar controls and configurable grammar structures.  ...  The semantic grammars are used by robust understanding technologies [Allen et al. (1996) ; Wang (1999) ; Bangalore and Johnston (2004) ] to map input utterances to the corresponding semantic representations  ... 
doi:10.1016/j.specom.2005.07.001 fatcat:2j2q3ownivdk5ipuoorucvmrcy

Re-evaluation of LVQ-HMM hybrid algorithm

Hitoshi Iwamida, Shigeru Katagiri, Erik McDermott
1993 Journal of the Acoustical Society of Japan (E)  
The LVQ-HMM hybrid algorithm was one of the first algorithms proposed in a recent approach aiming to integrate a highly discriminative artificial neural network-based classifier with an HMM capable of  ...  However, possible contribu tions and aspects of the algorithm needing further improvement are brought to light.  ...  Toshiyuki Hana zawa of the ATR Interpreting Telephony Research Laboratories (currently Mitsubishi Electrical Cor poration)for his help in using the HMM-LR soft -ware.  ... 
doi:10.1250/ast.14.267 fatcat:5lvmscxmk5grdgw3y4nqd2unr4

Page 356 of Computational Linguistics Vol. 26, Issue 3 [page]

2000 Computational Linguistics  
We tested various neural network models on the same six-class downsampled data used for decision tree training, using a variety of network architectures and out- put layer functions.  ...  The system detects sequences of distinctive pitch patterns by training one continuous- density HMM for each DA type.  ... 
« Previous Showing results 1 — 15 out of 2,189 results