Neural Networks, Fuzzy and Evolutionary Methods 2
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Modular Neural Networks for Low-Complex Phoneme Recognition
Authors:
Axel Glaeser, Ascom Systec Ltd. (Switzerland)
Page (NA) Paper number 133
Abstract:
We present a Modular Neural Network (MNN) for phoneme recognition within the framework of a hybrid system (neural networks and HMMs) for speakerindependent single word recognition. With this approach, we are taking the computational effort into account which is used as an additional criterion for assessing the system performance. The main idea of the proposed MNN is the distribution of the complexity for the phoneme classification task on a set of modules. Each of these modules is a single neural network which is characterized by its high degree of specialization. The number of interfaces, and therewith the possibilities for infiltering external acoustic-phonetic knowledge, increases for a modular architecture. Moreover, after the development of a suitable topology for the MNN, each of the modules can be optimized for its specific phoneme recognition task. This is done by detecting and pruning irrelevant input parameters and leads to a more efficient system in terms of memory and computational requirements.
SL980133.PDF (From Author)
SL980133.PDF (Rasterized)
TOP
Global Optimisation of Neural Network Models Via Sequential Sampling-Importance Resampling
Authors:
João F.G. de Freitas, Cambridge University (U.K.)
Sue E. Johnson, Cambridge University (U.K.)
Mahesan Niranjan, Cambridge University (U.K.)
Andrew H. Gee, Cambridge University (U.K.)
Page (NA) Paper number 213
Abstract:
We propose a novel strategy for training neural networks using sequential Monte Carlo algorithms. This global optimisation strategy allows us to learn the probability distribution of the network weights in a sequential framework. It is well suited to applications involving on-line, nonlinear or non-stationary signal processing. We show how the new algorithms can outperform extended Kalman filter (EKF) training.
SL980213.PDF (From Author)
SL980213.PDF (Rasterized)
TOP
Efficient Computation of MMI Neural Networks for Large Vocabulary Speech Recognition Systems
Authors:
Jörg Rottland, Duisburg University (Germany)
Andre Lüdecke, Duisburg University (Germany)
Gerhard Rigoll, Duisburg University (Germany)
Page (NA) Paper number 331
Abstract:
This paper describes, how to train Maximum Mutual Information Neural Networks (MMINN) in an efficient way, with a new topology. Large vocabulary speech recognition systems, based on a Hybrid MMI/connectionist HMM combination, have shown good performance on several tasks (RM and WSJ). MMINNs are trained to maximize the mutual information between the index of the winning output neuron (Winner-Takes-All network) and the phonetical class of the corresponding acoustic frame. One major problem of MMI-neural networks is the high computational effort, which is needed for the training of the neural networks. The computational effort is proportional to the input and output size of the neural network and to the number of training samples. This paper shows two approaches, that demonstrate, how these long training times can be reduced with very low or even no loss in recognition accuracy. This is achieved by the use of phonetical knowledge, to build a network topology based on phonetical classes.
SL980331.PDF (From Author)
SL980331.PDF (Rasterized)
TOP
Modular Connectionist Systems for Identifying Complex Arabic Phonetic Features
Authors:
Sid-Ahmed Selouani, IE/Houari boumedienne University of science and technology (Algeria)
Jean Caelen, IMAG/UJF (France)
Page (NA) Paper number 358
Abstract:
This paper concerns the identification of Arabic macro-classes and phonetic features by systems using a hierarchy of neural networks. These systems are composed of sub-neural-networks (SNNs) carrying out binary discrimination sub-tasks. Two types of architecture are presented: serial structure of experts and parallel disposition of them. This mixture of experts is composed of typically time delay neural networks using a version of autoregressive backpropagation algorithm (AR-TDNN). These hierarchical configurations are confronted to a monolithic system using standard backpropagation learning procedure. The test database consists of 60 VCV utterances and 50 phrases pronounced by 6 Algerian native speakers. The parallel configuration achieved much fewer error rate (13% vs. 16% and 28%) than other architectures. The parallel mixture of experts is incorporated in a hybrid structure (HMM-SNN) in the order to enhance performances of standard HMMs. Identification results show that 10% reduction of error rate is obtained by the hybrid system.