The process which involves generation of human like voice by a machine is called speech synthe-sis. The developments in the fteld of speech synthesis is vast in international languages, but it is limited in Indian languages like Kannada. This work aims at de-velopment of such a system for Kannada language using Festival and Festvox. It is based on parametric analysis and models of speech features, particular to a language and speaker. The system is memoryless and dynamic, wherein only extracted features are stored but not recorded audio. The training process involves speech data acquisition, pre-processing, labelling using Baum-Welch Iteration, whereas testing process involves text analysis, text segmentation, speech synthesis and qual-ity enhancement using acoustic HMM model develop-ment. The quality of synthesis is 3.52 dB to 5.02 dB as measured by Mel-Cepstral Distortion (MCD) score.
Index Terms-Baum-Welch algorithm, Hidden Markov Model (HMM), Kannada, MCD score, Speech synthesis.I.
Text <span lang="EN-US">to speech (TTS) is a system that generates artificial speech from text input. The prosodic models used improve the quality of the synthesized speech especially naturalness and intelligibility. The prosody involves intonation, intonation refers to the variations in the pitch frequency (F0) with respect to time in an utterance. This work mainly concentrates on building feedback neural network model to predict F0 contour in the utterances using Fujisaki intonation model parameters as the input features to the network since the Fujisaki intonation model is data driven and not a rule based one. In this work we have built 4-layer feedback neural network in the festival framework. Finally, the synthetically generated Kannada speech using the neural network model, is compared for its performance with the classification and regression tree (CART) model and Tilt model. Database of simple declarative Kannada sentences created by Carnegie Mellon University have been deployed in this work. From the study it is very clear that F0 contours can be accurately predicted using CART and neural network models, whereas naturalness and intelligibility is high in CART model rather than neural network model.</span>
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.