Stochastic Pronunciation Modeling by Ergodic-HMM of Acoustic Sub-word Units Export

@inproceedings{citeulike:326524, abstract = {We propose a stochastic pronunciation model using an ergodic - hidden Markov model (EHMM) of automatically derived acoustic sub-word units (SWU). The proposed EHMM discovers the pronunciation structure inherent in the acoustic training data of a word without any apriori phonetic transcriptions. The EHMM is an HMM of HMMs its states are SWU HMMs and the state-transitions compose various possible lexicon. The EHMM parameters are estimated by an iterative segmental -means procedure which jointly optimizes the subword units (states) and the pronunciation structure parameters (state-transitions). The EHMM based pronunciation model is evaluated in an English isolated word recognition task with 70 speakers drawn from 8 highly different first languages. Results show that EHMM learns the lexicon distribution over the population of speakers for each word, thereby effectively modeling the inter-speaker pronunciation variability. EHMM offers an improvement of 8\% (absolute) word recognition accuracy over a single most likely lexicon performance.}, author = {Ramasubramanian, V. and Srinivas, P. and Sreenivas, T. V.}, citeulike-article-id = {326524}, comment = {Aim is more robust detection of non-native speech (tolerance of mispronunciations). Achieved by better phone models trained using EHMM}, institution = {Department of Electrical Communication Engineering Indian Institute of Science}, keywords = {asr, eurospeech, hmm, l2a, machine\_learning, mathy, mispronunciation\_tolerance, modeling, notuseful}, month = {September}, organization = {Eurospeech}, posted-at = {2005-09-19 22:35:28}, priority = {0}, title = {Stochastic Pronunciation Modeling by Ergodic-HMM of Acoustic Sub-word Units}, year = {2005} }

TY - CONF ID - citeulike:326524 L3 - citeulike-article-id:326524 N2 - We propose a stochastic pronunciation model using an ergodic - hidden Markov model (EHMM) of automatically derived acoustic sub-word units (SWU). The proposed EHMM discovers the pronunciation structure inherent in the acoustic training data of a word without any apriori phonetic transcriptions. The EHMM is an HMM of HMMs its states are SWU HMMs and the state-transitions compose various possible lexicon. The EHMM parameters are estimated by an iterative segmental -means procedure which jointly optimizes the subword units (states) and the pronunciation structure parameters (state-transitions). The EHMM based pronunciation model is evaluated in an English isolated word recognition task with 70 speakers drawn from 8 highly different first languages. Results show that EHMM learns the lexicon distribution over the population of speakers for each word, thereby effectively modeling the inter-speaker pronunciation variability. EHMM offers an improvement of 8% (absolute) word recognition accuracy over a single most likely lexicon performance. OR - Eurospeech TI - Stochastic Pronunciation Modeling by Ergodic-HMM of Acoustic Sub-word Units KW - asr KW - eurospeech KW - hmm KW - l2a KW - machine_learning KW - mathy KW - mispronunciation_tolerance KW - modeling KW - notuseful N1 - Aim is more robust detection of non-native speech (tolerance of mispronunciations). Achieved by better phone models trained using EHMM AU - Ramasubramanian, V AU - Srinivas, P AU - Sreenivas, TV PY - 2005/09// ER -