Operational and experimental French telecommunication services using CNET speech recognition and text-to-speech synthesis

Sorin, Christel; Jouvet, Denis; Gagnoulet, C.; Dubois, D.; Sadek, David; Toularhoat, M.

doi:10.1016/0167-6393(95)00035-m

Cited by 22 publications

(6 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed algorithm was implemented with the CNET speech recognition softwares [13], [27], using the stateoptimized likelihood as performance criterion (segmental training). The acoustical parameters consisted of the logenergy plus the first eight mel cepstral coefficients with firstand second-order parameters appended.…”

Section: Resultsmentioning

confidence: 99%

“…If we look at this question from a more practical point of view, it is easily checked from the reestimation equations (23)- (27), that if two densities belonging to the same state are tied at some iteration, they will stay tied together for all subsequent iterations. Such a tying is thus equivalent to reducing by one the number of components in the considered mixture.…”

Section: Appendix C Extension To the Mixture Casementioning

confidence: 99%

See 1 more Smart Citation

An algorithm for maximum likelihood estimation of hidden Markov models with unknown state-tying

Cappé¹,

Mokbel²,

Jouvet³

et al. 1998

IEEE Trans. Speech Audio Process.

Self Cite

View full text Add to dashboard Cite

For speech recognition based on hidden Markov modeling, parameter-tying, which consists in constraining some of the parameters of the model to share the same value, has emerged as a standard practice. In this paper, an original algorithm is proposed that makes it possible to jointly estimate both the shared model parameters and the tying characteristics, using the maximum likelihood criterion. The proposed algorithm is based on a recently introduced extension of the classic expectationmaximization (EM) framework. The convergence properties of this class of algorithms are analyzed in detail. The method is evaluated on an isolated word recognition task using hidden Markov models (HMM's) with Gaussian observation densities and tying at the state level. Finally, the extension of this method to the case of mixture observation densities with tying at the mixture component level is discussed.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Appendix C Extension To the Mixture Casementioning

confidence: 99%

An algorithm for maximum likelihood estimation of hidden Markov models with unknown state-tying

Cappé¹,

Mokbel²,

Jouvet³

et al. 1998

IEEE Trans. Speech Audio Process.

Self Cite

View full text Add to dashboard Cite

show abstract

“…CNET Network Inc. (Center of the National d'Etudes des Telecommunication) is a telecommunication research and development center of France Telecom issued several criteria as a reference in the development of web-based application systems [25]. The criteria are as follows.…”

Section: Cnet (Builder)mentioning

confidence: 99%

A comparative analysis: Web design of Universities in Indonesian and American Countries based on CNET and Hall theory

Majid¹,

Ridwan²,

Setiadi³

et al. 2019

Proceedings of the International Conference of Social Science

View full text Add to dashboard Cite

The development of website-based technology can be useful in education. Universities can use web-based technology to communicate information to various fields without limitation, space and time. Universitas Pendidikan Indonesia (UPI) and American University (AU) are one of the universities that use website-based media to communicate information. The analytical methods used in the comparison are using CNET (builder) and Edward T Hall theory. Components in CNET (builder) include usability, a navigation system, graphic design, content, compatibility, loading time, function, while Hall components include animation, collectivism or individualism, product promotion values, transparency, linear vs. parallel navigation. The results are web AU looks simple with little image and content displayed. Many of the content is displayed on the main page on the web UPI, while at AU is not much; Images displayed on the web UPI further highlights the collectivity, while AU highlight individualists.

show abstract

“…The top 5 PTS programs available today are: Balabolka (Andipurnama, Wiguna, Susetyo, & Novianti, 2022;Pramono, 2022), NaturalReader (Fitria, 2022;Virdaus & Rifa'i, 2021), Lecteur de rêves vocal (Fichten et al, 2022;Fitria, 2022)., Amazon Polly (Carvalho Afonso, Almeida, Beça, Silva, & Covalenco, 2022;Chan, Ghosh, Rastrow, & Hoffmeister, 2023) and Speechify (Amin, 2022;Kiseleva & Novikov, 2023). Among the applications of text-to-speech in the field of telecommunications services (Sorin et al, 1995), we can mention an important aspect of the applications of text-to-speech synthesis concerns services for people with disabilities. For example, the development of a voice communication support system for a patient living in an Algerian environment could involve coding, compression and synthesis systems, in particular that carried out from the so-called TTS.…”

Section: Introductionmentioning

confidence: 99%

Symmetric Technique of Formant Transitions Generation for Using in Speech Synthesis of Arabic Language

Chegrani,

Guerti,

Boudraa

2024

Preprint

View full text Add to dashboard Cite

The objective of this work is the elaboration of a vocal communication aid system for the blind in the Arab world, which is at the same time used as a system for learning Arabic for them or for non-native speakers. This system is a result of the development of a speech synthesizer for the Arabic language which is based on the concatenative synthesis method (end-to-end speech data collection (Eskenazi, Levow, Meng, Parent, & Suendermann, 2013)) of sub-syllable sound units typically stored in digital wave format (Vaseghi, 2008), obtained using software called PRAAT (Gold, Morgan, & Ellis, 2011; Pleva, Juhár, & Thiessen, 2015)) from a natural language corpus (Chou, Tseng, & Lee, 2002). The treatments in this study took place in different phases (العماري, 2021). Text analysis techniques are responsible for converting incoming text into a linguistic representation that encodes information about how the input text should be pronounced (Shiga, Ni, Tachibana, & Okamoto, 2020). The method used to produce missing phonetic units in syllable counts (CV) is to use mathematical symmetry from (VC), which as a result obtained the construction of a phonetic lexicon based on a reduced number of phonemes, which is 124 units; And improving the patterns resulting from inserting formula transitions into (VC) generated from their counterpart (CV) using this feature. Finally, the implications of this after including the linear smoothing stage (Dutoit, 1994) are to obtain speech outputs that resemble natural speech as much as possible without noise (Millstein, 2020), continuous, and intelligible at different levels of language that are able to be employed in synthesising intelligible speech, for blind individuals in the Arab community, noting that this method is for all languages of the world in terms of producing the quality of speech.

show abstract

Operational and experimental French telecommunication services using CNET speech recognition and text-to-speech synthesis

Cited by 22 publications

References 11 publications

An algorithm for maximum likelihood estimation of hidden Markov models with unknown state-tying

An algorithm for maximum likelihood estimation of hidden Markov models with unknown state-tying

A comparative analysis: Web design of Universities in Indonesian and American Countries based on CNET and Hall theory

Symmetric Technique of Formant Transitions Generation for Using in Speech Synthesis of Arabic Language

Contact Info

Product

Resources

About