This paper presents a preliminary analysis of differences of formants among Shanghai, Guangzhou, Chongqing and Xiamen dialectal Putonghua (PTH) in China. An improved formant model based on the linear prediction (LP) feature analysis and a two-dimensional(2D) hidden Markov model (HMM) of formants is employed for estimation of the formant frequencies of vowels and diphthongs. The experiments show Guangzhou dialectal PTH (GZPTH) has the highest average frequency in 1st formant(F1) while Shanghai PTH(SHPTH) has lowest of F1 among four dialectal PTHs. In contrast, Chongqing PTH (CQPTH) gives highest frequency in 2nd and 3rd formants (F2, F3) with a comparison of other dialectal PTHs. The formant spaces of four dialectal PTHs have also been investigated.
展开▼