Sinewave Synthesis: Information
The SineWave Synthesis (SWS) system was developed at Haskins Laboratories by Philip Rubin in the late 1970s. This interactive demo of SWS was originally based on a HyperCard stack (HyperSWS) created by Rubin at the suggestion of Carol Fowler, then President of Haskins Laboratories, to demonstrate the phenomenon of sine-wave replication of speech at the Haskins Laboratories Board of Trustees Meeting, on Nov. 4, 1992. An additional inspiration for the original HyperSWS project was to illustrate this form of synthesis at the meeting of the Psychonomic Society in St. Louis on November 13, 1992. This interactive web demo focuses on the use of this program in the research of Rubin, Robert Remez, and their colleagues.
Technical information. The SineWave Synthesis program (SWS) is a software sound synthesizer intended to be used to create replicas of natural speech utterances by replacing the normal formant structure with time-varying pure tones (sinewaves). The synthesizer is implemented entirely in software, originally in Fortran, created on a Digital Equipment Corporation PDP-11/45 minicomputer. and then on the VAX series of computers running VMS. The sounds included here were extracted from the original HyperCard stack. The original method used to make the sounds and get them into HyperSWS was as follows: Formant center frequency and amplitude values were derived from natural utterances using both the ILS system (for Linear Predictive Coding, LPC, analysis) and the HADES program (for both LPC and FFT analyses). These formant and amplitude values were converted into input (.SWI) parameters for SWS. Audio files were then generated in Haskins (.PCM) format on the VAX and transferred from the VAX to a Macintosh in binary format. A commercial Macintosh program called Sound Designer II was used to “resample” the files (convert the sampling rate) from 10000 Hz (the Haskins PCM rate) to 11128 Hz (a Mac-compatible sampling rate). A commercial Macintosh speech manipulation/analysis program called Signalyze was used to open these files, edit them if necessary, add a DC offset of 128, and then save them in SoundEdit format. A commercial program called SoundEdit Pro was then used to low-pass filter and amplify the files, if necessary, and save them in the HyperSWS stack as resources. For this web version at that time, SoundEdit Pro was used again to open HyperSWS, extract the sound resources, and convert them into AIFF format.
SWS Instruction Manual. Version 2.1, VAX version, August 11, 1982. See: Rubin, P. E. (1980). Sinewave synthesis. Internal memorandum. Haskins Laboratories, New Haven, Connecticut.
Download all SWS input parameter files (.SWI) as a Zip file.
Download all audio files as a Zip in: MP3 format or WAV format.
SWS Overview and History, 2005. An unpublished draft from 2005 by Philip Rubin covering how the Haskins sinewave synthesis program was developed and used during its first 25 years. Includes early Fortran source code and an early version of MATLAB code from Dan Ellis. See below for a more current version of his code.
MATLAB routines: Dan Ellis (Previously: Professor, Electrical Engineering, Columbia University; and Currently: Research Scientist, Google, Inc., New York City) created Matlab routines that contain all the data needed to resynthesize sinewave speech. See: D. P. W. Ellis. (2004). "Sinewave Speech Analysis/Synthesis in Matlab", Web resource, https://www.ee.columbia.edu/~dpwe/resources/matlab/sws/.
Praat scripts. Sine-wave speech Praat scripts, by Chris Darwin
_____
Acknowledgments. The sinewave synthesis program (SWS) was developed by Philip Rubin at Haskins Laboratories in the late 1970's. Others who contributed to the development of the synthesizer include Robert Remez, Lenny Szubowicz and Michael D'Angelo. The original inspiration for work on this synthesizer came from the research of Peter Bailey, Quentin Summerfield, and Michael Dorman, and also work by Rod McGuire.
The research described here and the development of related stimuli were the collaborative effort of Robert Remez, Philip Rubin, and their colleagues: Stefanie M. Berns, Thomas D. Carrell, Jennifer M. Fellowes, William A. Howell, Jessica M. Lang, Elizabeth B. Lynch, Lynne C. Nygaard, Jennifer S. Pardo, David B. Pisoni, and Bella Schanzer. Mark Tiede provided expert advice on getting sound into the original HyperSWS stack.
Over the years many of our colleagues have contributed to our work by providing generous advice, technical assistance, stern encouragement, curiosity, expert criticism, and many valuable suggestions. This list includes: Arthur Abramson, Dick Aslin, Peter Bailey, Peter Balsam, Alan Bell, Eva Blumenthal, Al Bregman, Cathi Best, Jan Charles-Luce, Lisa Coady, Richard Crane, James E. Cutting, Chris Darwin, Kraig Eisenman, Dan Ellis, Carol Fowler, Lou Gerstman, Louis Goldstein, Vince Gracco, Stephen Grossberg, Donald Hailey, Arthur House, Julian Hochberg, Jim Jenkins, Mari Jones, Peter Jusczyk, J. A. Scott Kelso, Dennis Klatt, Carol Krumhansl, Peter Ladefoged, Harlan Lane, Alvin Liberman, Philip Lieberman, Leigh Lisker, Paul Luce, Bill Mace, Yvonne Manning-Jones, Charles Marshall, Ignatius Mattingly, Richard McGowan, Howard Nusbaum, Dick Pastore, Paula Payton, Jan Rabinowitz, Tim Rand, Bruno Repp, Frank Restle, Elliot Saltzman, Arthur Samuel, Jim Sawusch, Amanda Steinberg, Winifred Strange, Michael Studdert-Kennedy, Quentin Summerfield, Michael Turvey, Eric Vatikiotis-Bateson, Douglas H. Whalen, Ed Wiley, and Ress Young.
For their unstinting support and ability to tolerate silly sounds, we thank Harriet Greisser, Joette Katz, Jason Rubin and Samantha Katz.
This research has been supported by grants to Robert E. Remez (NIMH MH 3248; NICHD HD-15672; NIH NIDCD DC00308; NINDS NS-22096); Haskins Laboratories (NICHHD HD-01994), and David B. Pisoni (NIMH MH 24027).
