STRAIGHT information (This page is obsolete. New page is here. )

[Wakayama University]

STRAIGHT is a versatile speech manipulation tool invented by Hideki Kawahara when he was in ATR. A series of refinements and developments were conducted in the "Auditory Brain Project" under CREST program sponsored by JST.

This page on CD-ROM is modified to meet educational requirements for EDUCATION ARENA of Eurospeech'03 Geneve.

Visualizations of underlying principles

STRAIGHT is based on a simple channel VOCOER. It decomposes input speech signals into source parameters and spectral parameters. Successive refinements on each parameter extraction procedure enable the total system to resynthesize high-quality speech. These refinements provide educational case studies on how abstract signal processing concepts are applied to procees real world signals. (Note: Visualizations are encoded in DivX format. Please check the following URLs for players/decoders or a list ofsoftwares.)

Extended pitch synchronous analysis (visualization: 115MB)

Usual spectral analysis methods fail to represent smooth nature of voiced sounds because of interferences due to periodic excitation. This visualization illustrates how pitch synchronous analysis is extended to selectively eliminate the periodicity related interferences.

Instantaneous frequency and its use in wavelet based fundamental frequency estimation (visualization: 121MB)

Usual definition of period is not relevant for signals with changing periodicity like speech signals. Instantaneous frequency is an extension of frequency concept defined as the speed of phase change. This visualization illustrates how instantaneous frequency concept is relating to sinusoidal components of complex sounds and how wavelet analysis provides a mean to uniquely extract fundamental component. By definition, the instanteneous frequency of the fundamental component is the fundamental frequency (F0) of the speech signal under study.

Group delay and minimum phase component for representing source characteristics (visualization: 143MB)

Usual phase representation is sometimes misleading for understanding source characteristics. Group delay defined as a delivertive on the frequency axis is more intiutive representation by taking it as a centroid of energy at each frequency. Combining this interpretation with a concept of minimum phase concempt provides a very instructive interpretation of speech excitation. It illustrates when vocal tract excitation takes place at each frequency in a visually intuitive manner. Excitation source signal of STRAIGHT is designed based on this interpretation and approximation.

Demonstrations

Affine transformation

The basic application of STRAIGHT is to modify fundamental frequencies, the frequency axis and the time axis of a speech sample independently in a proportional fashion. The following examples illustrates results of these basic manipulations.

• Japanese sentence (spoken by a female speaker)[original]

  Table 1.

  	50% of the original F0	with original F0	200% of the original F0
  with frequency axis stretch (1.25)	sample 1	sample 2	sample 3
  without frequency axis modification	sample 4	sample 5	sample 6
  with frequency axis compression (0.8)	sample 7	sample 8	sample 9

Arbitrary transformation

Transformations applicable to STRAIGHT parameters are not necessarily proportional. Nonlinear and non-stationary transformations of parameters are allowed unless they do not violate physical feasibility of the modified representations.

• Artificial quintet made from two singing samples.
  [Sample singing (male)], [Sample singing (female)], [Artificial quintet]

Auditory morphing of speech sounds

Auditory morphing is to transform one speech example into the other speech example in a parameterized manner.

• From a male speech to a female speech. [Morphing rate -0.1 to 1.1 in 0.1 steps.]
• From "right" to "light". [Movie with time-frequency representations: Morphing rate -0.2 to 1.2 in 0.2 steps.]

References

• Hideki Kawahara, Ikuyo Masuda-Katsuse and Alain de Cheveigne: Restructuring speech representations using a pitch-adaptive time-frequency smoothing and an instantaneous-frequency-based F0 extraction: Possible role of a repetitive structure in sounds, Speech Communication, 27, 3-4, pp.187-207 (1999). [The EURASIP Best-Paper Award 1998/99] (draft)
• Hideki Kawahara, Haruhiro Katayose, Alain de Cheveigne, Roy D. Patterson: Fixed Point Analysis of Frequency to Instantaneous Frequency Mapping for Accurate Estimation of F0 and Periodicity , Proc. EUROSPEECH'99, Volume 6, Page 2781-2784 (1999). (PDF)
• Hideki Kawahara, Yoshinori Atake and Parham Zolfaghari: Accurate vocal event detection method based on a fixed-point to weighted average group delay, ICSLP-2000, Beijing, pp.664-667 2000. (PDF)
• Parham Zolfaghari, Yoshinori Atake, Kiyohiro Shikano, Hideki Kawahara: Investigation of analysis and synthesis parameters of STRAIGHT by subjective evaluation, ICSLP-2000, Beijin
• H. Kawahara and P. Zolfaghari: Systematic F0 glitches around vowel nasal transitions, EUROSPEECH'2001, pp.2459-2462, 2001.
• H. Kawahara, Jo Estill and O. Fujimura: Aperiodicity extraction and control using mixed mode excitation and group delay manipulation for a high quality speech analysis, modification and synthesis system STRAIGHT, MAVEBA 2001, Sept.13-15, Firentze Italy, 2001. (PDF)
• Hisami Matsui and Hideki Kawahara: Auditorily motivated elastic spectral distance and its application to emotional morphing of portrayal speech, FIRST PAN-AMERICAN/IBERIAN MEETING ON ACOUSTICS, 2-6 December 2002, Cancun, 3pSC11.
• Hideki Kawahara and Hisami Matsui: AUDITORY MORPHING BASED ON AN ELASTIC PERCEPTUAL DISTANCE METRIC IN AN INTERFERENCE-FREE TIME-FREQUENCY REPRESENTATION, Proc. ICASSP'2003, vol.I, pp.256-259, 2003. (PDF)
• Hideki Kawahara: Exemplar-based Voice Quality Analysis and Control using a High Quality Auditory Morphing Procedure based on STRAIGHT, VOQUAL'03, ISCA Tutorial and Research Workshop, Geneva, August 27-29, 2003, pp.109-114. (PDF)

Notes on availability

Essential components of STRAIGHT are patented by ATR and JST. When a company is interested in using STRAIGHT as a research tool, a written agreement between JST and the company is necessary for the company to get access to the codes and the technical documents. (Please contact the JST office for details.) For non-commercial research and/or educational institutes, please contact the web-master.