The Research On Open-loop Pitch Analysis Algorithm Based On Normalization Of Autocorrelation Function

At present, the speech quality in transmission of the channels and the coding rate isan increasing demand. At the same time, the increasingly strong demand for real-timetransmission of the speech, especially in some situations requires a higher property ofthe real-time, such as the transmission environment of the undersea, space, is morepressing to reduce the transmission and computation delay of the speech signal. Toreduce the transmission and computation delay there are two research directions: thefirst direction is lower the coding rate of the speech encoder at the expense ofsacrificing speech quality to reduce the same length of the speech transmission delay.The second direction is in the case of keeping a certain speech coding rate as far aspossible to reduce the computation, and to achieve real-time requirements by reducingthe computation delay.With the continuous development of the speech coding technology, more and morelow encoding rete speech coding standards have been proposed, such as used in IPtelephony, wireless communications, as well as a dedicated digital channel speechcoding standard G.729and its Appendix A, that is,G.729A is widely used. Low-ratespeech coding to very low bit rate speech coding, the research focus has shifted to lowerspeech coding rate on the complexity of the algorithm to reduce the algorithm of theG.729speech coding standard version, the voice quality is slightly decreased but thecomputation is reduced by about half. Nevertheless, in the pursuit of speech quality atthe same time, some of the hardware platform real-time requirements of the occasion isstill unacceptable. Therefore, reducing the G.729and the G.729A algorithmcomputation has been researched extensively.The paper focus on the part of the encoding process the large amount of calculationalgorithms, such as the calculation of the normalized autocorrelation function and theopen-loop pitch analysis, and an improved open-loop pitch analysis algorithm, to reducethe computational complexity of the algorithm. The paper use VC++6.0and MATLAB7.1carry out co-simulation, to verify its reconstruction of voice quality and encodingtime than the original coding algorithm is applied in C source code to the ITU-T, and inthe present research popular areas of the FPGA platform, but also improved after theopen-loop pitch analysis algorithm part of the VHDL source code, so that real-timeimplementation in hardware.Experimental data indicate that in the circumstances to ensure voice quality, improved speech encoder computation reduces by about10.8%, the basic guarantee forthe hardware implementation of some real-time requirements.