Sinteza govora z uporabo prikritih Markovovih modelov
نویسندگان
چکیده
Povzetek V tem članku je predstavljen postopek sinteze govora s pomočjo prikritih Markovovih modelov. Osnovo za pridobivanje značilk, s katerimi učimo modele, predstavlja vir–filter model govora. Vektor značilk tako sestavimo iz dveh delov. Prvi del opisuje govorni trakt, drugi del pa vzbujanje. Vhod v postopek sinteze predstavlja niz fonemov s pripadajočimi trajanji in poteki osnovne frekvence. Govor nato tvorimo tako, da iz ustreznih prikritih Markovovih modelov generiramo najverjetnej ši niz vektorjev značilk, iz katerega nato rekonstruiramo govor. Če vektor značilk razširimo še z dinamičnimi značilkami, dobimo gladke prehode med posameznimi glasovi.
منابع مشابه
Umetno tvorjenje slovenskega govora s pomočjo odprto kodnih orodij ter prikritih Markovovih modelov
Povzetek Na področju tvorjenja umetnega govora (sinteze) se je do danes uveljavilo veliko prosto dostopnih orodji, ki omogočajo gradnjo sistemov za umetno tvorjenje govora (sintetizatorjev). Programska orodja omogočajo s pomočjo vgrajenih uveljavljenih postopkov na uporabniku prijazen način udejanjiti lasten sintetizator govora. Ker so takšni sistemi močno odvisni od jezika, pri tovrstnih orodj...
متن کاملContext-Dependent Acoustic Modelling of Croatian Speech
This paper presents experiments of Croatian speech modelling used in speech recognition as well as in speech synthesis. The proposed acoustic model is based on context-dependent triphone hidden Markov models and Croatian phonetic rules. For speech recognition and speech synthesis experiments a common Croatian speech corpora is used. The experiments have shown that Croatian speech corpora, Croat...
متن کاملSpeech Synthesis and Discourse Information
This paper describes some recent work towards a conversational speech synthesis system for use in interactive dialogues between a human and an information system, robot, or speech translation device. The paper describes several response-type utterances that are currently very difficult to implement using traditional speech synthesis methods, and shows how these non-verbal speech sounds function...
متن کاملOptimization of Latent Semantic Analysis based Language Model Interpolation for Meeting Recognition
Latent Semantic Analysis (LSA) defines a semantic similarity space using a training corpus. This semantic similarity can be used for dealing with long distance dependencies, which are an inherent problem for traditional word-based n-gram models. This paper presents an analysis of interpolated LSA models that are applied to meeting recognition. For this task it is necessary to combine meeting an...
متن کاملInfluence of Mosfet Model Form on Characteristics of the Boost Converter
In the paper boost converter characteristics at the steady state obtained from SPICE analysis with the use of selected kinds of MOSFET models of various complexity and accuracy are compared. The dependencies of the converter output voltage, the watt-hour efficiency and the MOSFET inner temperature on the frequency and the duty cycle of the MOSFET control signal as well as the converter load res...
متن کامل