Isarn phoneme transcription using statistical model and transcription rule

Janyoi, Pongsathon; Seresangtakul, Pusadee

doi:10.2495/icacc130461

Cited by 4 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Three main types of approaches have been proposed for successful word segmentation: rule-based, statistical-based, and hybrid approaches. Efforts in research proposing word segmentation for other languages [13][14][15], has suggested that the statistic-based approach achieves the best possible performance. However, the training model requires a fairly large text corpus.…”

Section: Word Segmentationmentioning

confidence: 99%

“…For example, the minus symbol ("-") is pronounced as "ลบ" /lop/ (minus) when it occurs between two numbers in a mathematical expression, and as "ฮอด" /hɔːt/ (to) when it represents a period of time. The readable text produced is then converted into the linguistic specification, t hrough a h ybrid of the statistical model and rule-based approaches [15].…”

Section: Phonetic Transcriptionmentioning

confidence: 99%

“…This paper proposes herein the development of an HMM-based speech synthesis system for the Isarn language, which is classified as a low, limited resource language. In previous work, we proposed a text analysis module for converting the input text to the corresponding linguistic specification [15]. We also proposed the Isarn HMM-based speech synthesis [16].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Isarn Dialect Speech Synthesis using HMM with syllable-context features

Janyoi

Seresangtakul

2018

ECTI-CIT

Self Cite

View full text Add to dashboard Cite

This paper describes the Isarn speech synthesis system, which is a regional dialect spoken in the Northeast of Thailand. In this study, we focus to improve the prosody generation of the system by using the additional context features. In order to develop the system, the speech parameters (Mel-ceptrum and fundamental frequencies of phoneme within different phonetic contexts) were modelled using Hidden Markov Models (HMM). Synthetic speech was generated by converting the input text into context-dependent phonemes. Speech parameters were generated from the trained HMM, according to the context-dependent phonemes, and were then synthesized through a speech vocoder. In this study, systems were trained using three different feature sets: basic contextual features, tonal, and syllable-context features. Objective and subjective tests were conducted to determine the performance of the proposed system. The results indicated that the addition of the syllable-context features significantly improved the naturalness of synthesized speech.

show abstract

Section: Word Segmentationmentioning

confidence: 99%

Section: Phonetic Transcriptionmentioning

confidence: 99%