Abstract—This work aims to find the most effective method for natural and good sound quality, after a comparative evaluation, the best method approved by this evaluation is used in our HTS_ARAB_TALK system. HTS is a system speech synthesis based on HMM, which is a new technique relative to other synthesis techniques. Several versions of HMMs are developed, with varying contextual information, algorithms for estimating the parameters of the source-filter synthesis model and extract the coefficients aperiodicity if the STRAIGHT vocoder is used to extract the F0 and obtain the spectrum and autoregressive HMM model. These methods are compared, in a perceptive test, to the naturalness of speech. The evaluation shows that the use of STRAIGHT and MATLAB with HTS significantly improves synthesis naturalness compared to the state of the art.

 

Index Terms—hidden markov MODEL, autoregressive HMM, speech synthesis, Arabic language, HTS, HTS_ARAB_TALK

Cite: M. K. Krichi and A. Cherif, "An Evaluation of Techniques Based on HMM Speech Synthesis for Using in HTS-ARAB-TALK," International Journal of Signal Processing Systems, Vol. 4, No. 2, pp. 133-138, April 2016. doi: 10.12720/ijsps.4.2.133-138