Study On The Technology Of Pitch-Scale Modification

Along with the development of information technology and multimedia technology,people's leisure life is more and more colorful,the requirements of some audio materials are also more and more high,ordinary audio materials can not meet the needs of people's life.The pitch-scale modification technology emerges as the times require,it adjust the tone of given speech according to some algorithm,without changing the speech intelligibility and keeps the speed not change.In real life,it's widely used.For example,it can make a person's voice that you are familiar with convert into a voice that you are not familiar with which can protect the privacy of the individual.In addition,human voice be embellished is converted into animal sounds,so pitch-scale modification can reached the purpose of providing entertainment for people,etc.At present,there are many algorithms,which mainly divided into time domain algorithm and frequency domain algorithm.Time domain algorithm mainly includes time domain modulation and Synchronized Fixed Synthesis,the frequency domain algorithm mainly includes frequency domain interpolation and phase-vocoder algorithm.This paper introduces the principle of the existing pitch-scale modification algorithm,at the same time,the advantages and disadvantages of each method are also introduced,which is more convenient for people to choose different pitch-scale modification methods according to the specific circumstances.The SOLA-FS algorithm is widely used,and it is divided into two steps: sample rate conversion and time warping.For the original speech signal sampling point,we combined interpolation with decimation to realize sampling rate conversion.After sampling rate conversion,the length of the original speech become longer or shorter,if you want to keep the speed constant,you need to do time scale modification with SOLA-FS.The algorithm is widely used for its simplicity,but due to the segmentation processing of speech,phase discontinuity exists in each connection point,resulting in the voice with poor quality.For the problems of SOLA-FS algorithm,this paper improves it.The new algorithm does not change the original speech sampling rate during speech wave transform.But the modified speech will be played with a new sampling rate.The new algorithm realizes the sampling rate conversion through coping the lastpitch or deleting the last pitch of each frame.And the length of each frame of speech can get by the back stepping,the maximum correlation coefficient between the last pitch and a pitch period before it is obtained with the method of correlation function,in order to determine the optimal length of the copy or delete section of.The improvement of the above two points,greatly improved the phase discontinuity problems at connection point.Finally,MATLAB simulation experiments were designed to evaluate the new algorithm and to evaluate the effect of sound.Experimental results show that the new method improves the phase discontinuity problems at connection point obviously and gives a better tone effect compared with the SOLA-FS algorithm.