A comparative analysis on cepstrum, linear predictive coding and particle filtering based formant estimation methods

Formants are able to define basic properties of speech efficiently by using very limited parameter sets; thus they have found important usage area at many applications of speech processing like coding, recognition, synthesis and enhancement. Estimation of formants is harder than simply tracking the peaks of the spectrum; as the output of the vocal tract´s spectral peaks are dependent on the shape of vocal tract, excitation and periodicity in a complex way. Because of this reason, a lot of past work was done on formant estimation and their positive and negative properties have been recognized. In this article we analyzed some of the popular formant estimation method´s performances and compared them. Among these three compared methods, it´s seen that the particle filtering based formant estimation method gives the most successful performance. Furthermore, it´s recognized that linear predictive coding method has estimation difficulties with signals with low sampling frequencies and cepstrum method causes excess formants at peak picking.