Hidden Markov Model Techniques 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
Real-Time Probabilistic Segmentation for Segment-Based Speech Recognition
Authors:
Steven C. Lee, MIT Lab for Computer Science (USA)
James R. Glass, MIT Lab for Computer Science (USA)
Page (NA) Paper number 594
Abstract:
In this work, we investigate modifications to a probabilistic segmentation algorithm to achieve a real-time, and pipelined capability for a segment-based speech recognizer. The existing algorithm used a Viterbi and backwards A* search to hypothesize phonetic segments. We were able to reduce the computational requirements of this algorithm by reducing the effective search space to acoustic landmarks, and were able to achieve pipelined capability by executing the A* search in blocks defined by reliably detected phonetic boundaries. The new algorithm produces 30% fewer segments, and improves TIMIT phonetic recognition performance by 2.4% over an acoustic segmentation baseline. We were also able to produce 30% fewer segments on a word recognition task in a weather information domain.
SL980594.PDF (From Author)
SL980594.PDF (Rasterized)
TOP
Toward Markov Random Field Modeling of Speech
Authors:
Guillaume Gravier, ENST/TSi CNRS-URA 820 (France)
Marc Sigelle, ENST/TSI CBRS-URA 820 (France)
Gérard Chollet, ENST/TSI CNRS-URA 820 (France)
Page (NA) Paper number 560
Abstract:
In this paper, we present a new technique for statistical modeling of speech segments based on Markov random fields. Classical and multi-stream HMMs are particular cases of this more general family of models. However, the Random Field Model (RFM) proposed here can be seen as an extension of the multi-band HMM in which interactions between the frequency bands have been added. In a first experiment, samples are drawn from different models and compared to real observations. This experiment shows that the RFM is able to produce realistic samples but a single HMM still performs better. Isolated word recognition experiments stress the fact that more work must be done on the RFM in order to reach the performances of classical hidden Markov modeling techniques. For the moment, the RFM parameters are estimated using a heuristic. We believe that a real maximum likelihood parameter estimation algorithm should improve the results. The main advantage of this new model is that it can easily be extended since a model is defined by some local interactions and the Gibbs potential functions associated to those interactions.
SL980560.PDF (From Author)
SL980560.PDF (Rasterized)
TOP
Hidden Markov Models for Trajectory Modeling
Authors:
Rukmini Iyer, GTE/BBN Technologies (USA)
Herbert Gish, GTE/BBN Technologies (USA)
Man-Hung Siu, GTE/BBN Technologies (USA)
George Zavaliagkos, GTE/BBN Technologies (USA)
Spyros Matsoukas, GTE/BBN Technologies (USA)
Page (NA) Paper number 891
Abstract:
Current state-of-the-art statistical speech recognition systems use hidden Markov models (HMM) for modeling the speech signal. However, it is well known that HMM's do not exploit the time-dependence in the speech process, since they are limited by the assumption of conditional independence of observations given the state sequence. Alternative techniques, such as segment modeling approaches, can effectively exploit time-dependencies in the acoustic signal by discarding the observation independence assumption. However, losing the basic HMM structure is often a high computational price to pay for improved acoustic models. In this paper, we introduce the parallel path HMM that exploits the time-dependence in speech via parametric trajectory models while maintaining the HMM framework. We present preliminary results on Switchboard, a large vocabulary conversational speech recognition task, demonstrating both improved modeling and potential for improved recognition performance.