TANDEM-STRAIGHT is a speech analysis, modification, and synthesis system and widely used in the speech research community. It decomposes input speech into three types of parameters: Fundamental frequency, spectral envelope, and aperiodicity. Computational cost of TANDEM-STRAIGHT is lower than that of STRAIGHT which is a prior version of TANDEM-STRAIGHT. Sound quality of TANDEM-STRAIGHT is the same as that of STRAIGHT.

Speech processing by TANDEM-STRAIGHT

Figure illustrates the speech processing by TANDEM-STRAIGHT. TANDEM-STRAIGHT decompoases input speech into three parameters, and we can independently modify them and generate speech from three parameters. For example, we can manipulate the intonation by controlling F0 trajectory. Three parameters are effective as the parameters to analyze para- and non-linguistic information.