| With the development of information and multimedia technology, normal audio and video materials cannot satisfy people's specific requirement, such as for periodicity and entertainment. Then, the pitch shifting technology rise in response to the proper time and conditions. Pitch shifting is a method that not only alters the tone but also not change the duration according certain algorithm. Different tone is mainly embodied in different pitch period and formant frequency. So we can change the pitch period and formant frequency in order to achieve pitch shifting. The pitch shifting methods presented now always have kinds of defects.Speech synthesis technology is a kind of speech signal processing technology that developed with the requirement of human-computer interaction. Speech synthesis technology is a method that could convert the text to natural speech signal, which is generally called TTS. Early TTS mainly adopts parameter synthesis methods, such as formant synthesis method and LPC method. Parameter synthesis method is mature in theory and easy to be implemented, but the synthesized speech signal is not natural and has an obvious artificial feeling. In the early 1990s, the PSOLA method has been used in speech synthesis. Different with the traditional splicing method, the PSOLA method firstly has an analysis on the speech signal, in order to acquire the pitch marks, and then has a flexible adjustment on the prosody feathers, such as fundamental frequency, duration and intensity. Based on the study of kinds of speech synthesis methods and the PSOLA method, this paper develops a speech synthesis method that could deal with the pitch and duration information partly. Then, we have a simulation on the TD-PSOLA method and the speech synthesis method developed in this paper. According the results of the simulation, we find that the method developed in this paper is effective.The main works of this paper is as follow:1,In this paper, the presented pitch shifting methods are studied and implemented, especially the three typical pitch shifting methods, which are SOLA-FS method, interpolation-on-frequency method, and the phase vocoder method. Meanwhile, the merits and faults are given about them. Also, the author provides an improved pitch shifting methods based on SOLA-FS method, Trough the experiments, we find that the pitch-shifted audio signal not only alters the tone,but also not change the duration,and also gain certain improvement on the decreasing noises and the computation complexity. According to the simulation, we found that the improved SOLA-FS method could acquire more natural speech signal than other pitch shifting methods. Also, the Sound Quality Evaluating on three different methods is given in this paper. The results of auditory evaluating tests show that, under the given pitch-shifting ratio, whether up or down, the sound quality processed by the improved SOLA-FS method is the best of all.2,Combing the improved SOLA-FS method and the traditional PSOLA method, The author provides a new speech synthesis method,during which the pitch and the duration is dealt with partly. This method can not only keep the voice unit clear and natural, but also improve the capability of altering prosody. At end, the simulation results is given about TD-PSOLA method and the method provided in this paper. Under the different pitch scale ratio, the time domain chart and the pitch contour chart is given about the two speech synthesis method, also the comparison of the complexity of the two methods is given. According to the experiments, we find that the traditional TD-PSOLA method has a bigger difference from the envelope of the original pitch contour than the method provided in this paper, especially when the pitch scale ratio is bigger. But when the ratio is bigger, the new method can also gain better effect. Merely, the new method has a big complexity, but with the development of the computer and data store technology, the complexity would not be a question. |
PDF Full Text Request