Speech synthesis using neural networks

    公开(公告)号:GB2326321A

    公开(公告)日:1998-12-16

    申请号:GB9812479

    申请日:1998-06-11

    Applicant: MOTOROLA INC

    Abstract: A method is presented for providing, in response to a lexical pronunciation, efficient generation of a postlexical pronunciation, including the steps of: determining lexical phones, lexical features, and boundary information for a predetermined portion of text; and utilizing a pretrained neural network that was pretrained using lexical phones, postlexical phones, lexical features, and boundary information to generate a neural network hypothesis for a postlexical pronunciation of the predetermined portion of text.

    METHODE, DISPOSITIF ET SYSTEME POUR LA DESAMBIGUISATION DES PARTIES DU DISCOURS.

    公开(公告)号:BE1011964A3

    公开(公告)日:2000-03-07

    申请号:BE9800813

    申请日:1998-11-06

    Applicant: MOTOROLA INC

    Abstract: Une méthode (300), un dispositif (408) et système (400) fournissent une disambiguïsation des parties du discours pour des mots en se basant sur un traitement hybride stochastique et par réseau neural. La méthode désambiguïse les étiquettes des parties du discours de symboles de texte en obtenant un ensemble d'étiquettes annotées de manière probabiliste pour chaque symbole de texte, en déterminant un étiquette prévue localement pour chaque symbole de texte en se basant sur le contexte local du symbole de texte, en déterminant une étiquette de rechange pour chaque symbole de texte en se basant sur le contexte étendu du symbole de texte, et en choisissant entre l'étiquette prévue localement et l'étiquette de rechange sont différentes.

    System for animating virtual actors using linguistic representations of speech for visual realism.

    公开(公告)号:GB2328849A

    公开(公告)日:1999-03-03

    申请号:GB9815620

    申请日:1998-07-20

    Applicant: MOTOROLA INC

    Abstract: A neural network based parameter system 118 is used for generating a virtual actor (visually-rendered model with speech 132) of which the movements are correlated with synthetic speech.Text 102, used to drive the virtual actor, is converted 104 to a linguistic representation of speech 106, which is converted to neural network linguistic parameters 110 by pre-processor 108. The neural network module 112 converts the neural network linguistic parameters into raw spatial parameters 114, which are finally converted into model parameters 120 by a post-processor 116. These model parameters 120 are then used to drive the virtual actors. Alternatively, a non-neural network based linguistics-to-speech module is used to convert the linguistic representation of speech 106. The speaker profile of a linguistics-to-speech module 126 provides data to change the characteristics of the alternatively-synthesized speech 128.

Patent Agency Ranking