For Pitch Extraction

A synthesis-based method for pitch extraction of the speech signal is proposed. The method synthesizes a numb¢: of log power spectra for different values of fundamental frequency and compares them with the log power spectrum of the input speech segment. The average magnitude (AM) difference between the two spectra is used for comparison. The value of fundamental frequency that gives the minimum AM difference between the synthesized spectrum and the input spectrum is chosen as the estimated value of fundamental frequency. The voiced/unvoiced decision is made on the basis of the value of the AM difference at the minimum. For synthesizing the log power spectrum, the speech signal is assumed to be the output of an all-pole filter. The transfer function of the all-pole filter is estimated from the input speech segment by using the autocorrelation method of linear prediction. The synthesis-based method is tried out on real speech data and the results are discussed. R6sum6. Une m6thode d'extraction de ia fondamentale du signal de parole bas6e sur ia synth~se est proposb.e. Elle procb, de par la synth6se d'un certain nombre de spectres iogarithmiques de puissance pour diff6rentes valeurs de la fr6quence fondamentale, et pour ia comparaison de ces spectres avec le spectre logarithmique de puissance du segment de parole analys6. La diff6rence d'amplitude moyenne (AM) entre les deux spectres est utilis6e pour ia comparaison. La valeur de la fr6quence fondamentale qui fournit ie minimum de diff6rence AM entre le spectre synth6tis6 et ie spectre du signal d'entr6e est choisie comme estimation de ia hauteur. La d6cision vois6/non-voise est 6tablie sur la valeur de la diff6rence AM au minimum. Pour synth6tiser le spectre de puissance, le signal de parole est suppos6 repr6senter la sortie d'un filtre tout pble. La fonction de transfert du fiitre tout p61e est estim6e ~t partir du segment de parole par la m6thode d'autocorr61ation de ia pr6diction lin6aire. Cette m6thode bas6e sur la synth6se est test6e sur des segments de parole naturelle et les r6sultats sont discut6s. Introduction about speech segments and the pitch period (or its inverse, the fundamental frequency) is estimated Pitch extraction is an important problem in for the voiced speech segments. The pitch contours speech analysis, in pitch extraction of a speech of speech utterances are useful in various speech utterance, voiced/unvoiced decisions are made processing applications such as speech analysis-synthesis [!,2], speech understanding [31 and