The method of pitch frequency detection on the base of tuning to its harmonics

摘要

The major part of speech compression algorithms from the first simple formant vocoders to widely used at presence LP- and MBE-vocoders in spite of different principals of coding use by the analysis-synthesis procedures information about the frequency of pitch [1-3]. There is a number of methods for its detection [4-6] : on the base of auto-correlation processing, cepstrum analysis, after linear prediction remaining signal. The methods provide good results when analysing parts of speech with pronounced vocalisation by insignificant noise level. When there is a strong noise component in the input signal the number of errors and accuracy of pitch frequency detection is greatly increased [7]. To eliminate the errors they use generally either method of linear smoothing of pitch contour that additionally decrease the accuracy. The paper presents the method and algorithm of pitch frequency detection which doesn't provide gross errors and lias the high accuracy.