Subject: Speech Processing

Topic: Speech Processing Applications

Difficulty: Medium

(i) The synthetic speech is provide with the necessary spectral, envelope, matched formants without the formants being explicitly identified by the LPC filter.

(ii) Although this creates an intelligible speech but it does not produce a natural-sounding speech because of the time design of excitation model.

(iii) Despite the fact that this kind of LPC synthesis clearly does not sound natural, this model can be subjected to spectral modification based on geometry differences of vocal tract mostly the male-to-female speech conversion, due to its unique relationship to vocal tract geometry.

(iv) The modified version of LPC based encoder and decoders such as CELP now being using advanced version of LPC.

(v) Although both these methods allow for changes in spectral envelope, it is difficult to achieve fine control of spectral detail using this method.