Speech Prosody 2008
A speech analysis, modification, and resynthesis system called STRAIGHT has been widely used in the speech research community. However, its foundation and implementation were not well established. This lecture introduces recent advances in STRAIGHT’s foundation based on a new concept called TANDEM, a simple method for calculating temporally stable power spectra using two F0-adaptive time windows. A new interpretation of Shannon’s sampling theory also provided a mathematical foundation and guidelines for designing STRAIGHT. A unified approach based on TANDEM and STRAIGHT in F0 extraction (at the same time, revisiting its conceptualization) and aperiodicity information representation provide flexibility in prosodyrelated feature manipulation. An overview on prospective applications of the new TANDEM-STRAIGHT in L2 study will also be presented.
Bibliographic reference. Kawahara, Hideki (2008): "TANDEM-STRAIGHT, a research tool for L2 study enabling flexible manipulations of prosodic information", In SP-2008, 619-628.