DNN Based Myanmar Speech Synthesis

Hlaing, Aye Mya; Pa, Win Pa; Thu, Ye Kyaw

doi:10.21437/sltu.2018-30

Cited by 5 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Myanmar pronunciation lexicon with syllable information was prepared because syllable is the basic sound unit bearing tone information in Myanmar language [10]. Many contextual labels reported in [11] formatted as HTS-style labels 2 have been extacted for Myanmar language. A question set is used in linguistic feature extraction for DNN and LSTM-RNN based speech synthesis and it is also language dependent requirement.…”

Section: Linguistic Feature Extraction For Myanmar Languagementioning

confidence: 99%

“…There is no publicly available Myanmar question set for linguistic feature extraction. We proposed Myanmar question set in [11] and it was used for extracting linguistic features. In Myanmar language, tone is the integral part of the pronunciation of syllable and can affect the meaning of that syllable.…”

Section: Linguistic Feature Extraction For Myanmar Languagementioning

confidence: 99%

“…Merlin speech synthesis toolkit [16] with Keras [17] python library was applied for modeling all systems on K80 GPU for training. DNN-based speech synthesis [11] was used as the baseline in this paper.…”

Section: Experimental Setupsmentioning

confidence: 99%

“…Little research has been performed for speech synthesis on Myanmar language former known as Burmese. Only three SPSS based papers on Myanmar speech synthesis are found publicly: HMM-based Myanmar TTS [9], CART-based Myanmar TTS [10], and DNN-based Myanmar speech synthesis [11]. In [9], the first HMM-based Myanmar TTS was operated at the syllable level and word information was used in CART-based Myanmar TTS in [10].…”

Section: Introductionmentioning

confidence: 99%

“…In [9], the first HMM-based Myanmar TTS was operated at the syllable level and word information was used in CART-based Myanmar TTS in [10]. In [11], more contextual information was applied in Myanmar speech synthesis.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Enhancing Myanmar Speech Synthesis with Linguistic Information and LSTM-RNN

Hlaing¹,

Pa²,

Thu³

2019

10th ISCA Workshop on Speech Synthesis (SSW 10)

View full text Add to dashboard Cite

Recently, Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) has become an attractive architecture in speech synthesis for its ability to learn long time-dependencies. Contextual linguistic information is an important feature for naturalness in speech synthesis and using that feature in various speech synthesis models improves the quality of the synthesized speeches for languages. In this paper, LSTM-RNN was applied in Myanmar speech synthesis, and the importance of contextual linguistic features and the effect of applying explicit tone information in different architectures of LSTM-RNN was examined using our proposed Myanmar question set. Experiments of LSTM-RNN, and a hybrid system of DNN and LSTM-RNN, i.e., four feedforward hidden layers followed by two LSTM-RNN layers, were done on Myanmar speech synthesis and compared with the baseline DNN. Both objective and subjective evaluations show that the hybrid of DNN and LSTM-RNN system gives more satisfiable synthesized speeches for Myanmar language than the LSTM-RNN and baseline DNN systems.

show abstract

Section: Linguistic Feature Extraction For Myanmar Languagementioning

confidence: 99%

Section: Linguistic Feature Extraction For Myanmar Languagementioning

confidence: 99%