A popular model for the human vocal system is that the vocal cords act like a "repeated pulse source" that feed a pulse train into the vocal tract, which in turn acts as a filter to shape the sound. The period between the pulses emitted by the vocal cords determine the pitch.

Pitch is one of important voice characteristics of human. It could provide much information about the speaker, such as the gender, ages (adults or children). Another important trait of the speech is formant. So it is interesting to change the pitch and preserve the formant or the spectral envelope. We could shift the pitch of  a man’s voice ,and the resynthesized speech may be regarded as a woman’s voice due to the pitch shifted up by a factor of 2, and vice versa. At the same time , we need to preserve the vocal tract information and the tempo of the processed speech for the speaker. In contrast, we did the tempo changing (time stretching) but did not alter the pitch and formant of the speaker. This topic seems easy to do, however it involves a lot of techniques to achieve these goals, which is not just down-sampling and up-sampling or just change sampling frequency of the speech signals.

Our instructor, Prof. Harris taught us more about pitch related techniques and gave us a good opportunity to understand this issue in E-fair. so, we have a quit good understanding about this issue and got the right research topic.

We devoted to this projects for two weeks , and we complete this project . The result are desirable. We completed one software regarding to the pitch shifting and time stretching of human speeches. We achieve the pitch shifting up and down by factor of two , and also complete the time stretching part .

However, we meet two main problems, one is when we process the samples, i.e. the background

music is changed due to its own pitch is shifted up or down at the same time, which is not we want. The other is if we alter the pitch up by factor of 4 or more, the result is not satisfactory, because we can not preserve the important frequency components within the ranges of human hearing, which means lots of frequency information lost. Although the projects was completed as we expect, we need do more research on those important issues in the future study.