Effect of Linear Fitting Technique on Quality of Synthesized Speech

The evolutionary origins of speech remain obscure. This contribution presents a novel approach to smooth the human speech using poly fitting techniques. The acquisition of speech signal is carried out using high quality system. Acoustic analysis of Hindi language female/male speaker is done via signal processing technique. An algorithm is designed to modify various speech parameters with different frame size. Various frame level ranging from 10 to 50 are taken and linear fitting of the data is carried out. Perceptual experiments were also conducted to assess the quality of the synthesized speech. Synthesized female speech showed low value of PESQ score with respect to male speech.