Speech Proc Lab

Current Projects, Sponsors, and Related Publications

Some of the current research activities in the Speech Processing Laboratory are described on this page:

National Gallery of the Spoken Word (digital libraries)- An ongoing 5 year project, The National Gallery of the Spoken Word (NGSW) is creating a significant, fully searchable online database of spoken word collections spanning the 20th century—the first large-scale repository of its kind. NGSW provides storage for these digital holdings and public exhibit "space" for the most evocative collections. From Thomas Edison's first cylinder recordings, to the voices of Babe Ruth and Florence Nightingale, and Studs Terkel's timeless interviews, the collections of the NGSW cover a variety of interests and topics. Designed as an expansive repository of aural resources, NGSW will grow to include many more collections from partnering institutions around the country. One of the primary goals of the NGSW is the development of a rich set of exhibits and educational curricula that fully incorporate sound files.

The NGSW project is a very large scale undertaking involving four colleges at MSU and numerous partner institutions across the country (see NGSW web page). The Speech Processing Laboratory has worked along two research tracks, the primary focus being the development of secure watermarking algorithms for NGSW copyright protection. The secondary effort involves the development support technologies for the robust audio search engine being developed at the Center for Spoken Language Research at the University of Colorado, Boulder.

Example publications:

A. Gurijala, J.R. Deller, Jr., M.S. Seadle, and J.H.L. Hansen, “Watermarking through parametric modeling,” Proc. Int'l. Conf. Spoken Language Processing, Denver, Sep. 2002 (in press).

J.H.L. Hansen, J.R. Deller, Jr., and M.S. Seadle, “Transcript-free search of audio archives for the National Gallery of the Spoken Word,” Proc. First ACM/IEEE Joint Conf. Digital Libraries, Roanoke, June 2001. (Published on CD-ROM)

Sponsors:

This project is sponsored by the Digital Libraries Initiative - Phase II, a consortium of federal agencies led by the U.S. National Science Foundation.

Language identification - The Speech Processing Lab is collaborating with researchers in the Information Systems Technology Group at MIT Lincoln Laboratory in the development of new algorithms for the identification of the natural language of a spoken utterance. This problem is of increasing importance in the communications technologies, and in numerous other commercial and military applications.

Example publications:

P.A. Torres-Carrasquillo, D.A. Reynolds, and J.R. Deller, Jr., “Language identification using Gaussian mixture model tokenization,” Proc. IEEE Int'l. Conf. Acoustics, Speech, and Signal Processing, Orlando, May 2002. (Published on CD-ROM)

P.A. Torres-Carrasquillo, D.A. Reynolds, E. Singer, M.A. Kohler, R.J. Greene, and J.R. Deller, Jr., “Approaches to language identification using Gaussian mixture models and shifted delta cepstral features,” Proc. Int'l. Conf. Spoken Language Processing, Denver, Sep. 2002 (in press).

Sponsors:

This work is receiving support from MIT Lincoln Laboratory, the GE Faculty Fellowship Program, and the Alfred P. Sloan Foundation.

Set-membership identification & filtering - This project involves a wide-ranging group of activities focusing upon development, theoretical analysis, and applications of set-theoretic signal processing. Recent areas of interest include convergence analysis of set-based identification algorithms for parametric models, and applications to adaptive tracking, pattern recognition, and speech modeling. Strong collaborations in this area exist among the Speech Processing Lab, and researchers at the University of Notre Dame and at Tulane University.

Example publications:

J.R. Deller, Jr. and Y.F. Huang, “Set-membership identification and filtering in signal processing,” invited paper in: Circuits, Systems, and Signal Processing [Special issue on signal processing and its applications, P. Stoica (section editor for estimation and identification)], to appear Dec. 2002.

J.R. Deller, Jr., S. Gollamudi, S. Nagaraj, and Y.-F. Huang, “Convergence of the Quasi-OBE algorithm: Proofs and performance implications” (invited paper) Proc. IFAC / IFORS 12th Int. Conf. on System Identification and Parameter Estimation (SYSID 2000) (published on CD-ROM), Santa Barbara, July 2000.

D. Joachim and J.R. Deller, Jr., “Short signal classification using set-membership identification: Application to speech labeling," Proc. 43rd Midwest Symp. Circuits and Systems (published on CD-ROM), East Lansing, Aug. 2000.

Sponsors:

These projects have received prior support from the National Science Foundation, the U.S. Office of Naval Research, and from a visiting faculty appointment at the University of Notre Dame, Department of Electrical Engineering.

Speech recognition for augmentative communications - The Speech Processing Lab is collaborating on several projects with InvoTek, Inc. of Alma, Arkansas in the creation of alternative and assistive communications devices for persons with motor and speech challenges. A focus of current work is the development of a "voice access system" that will allow persons with very limited speech to control a personal computer or other digital device using "speech" inputs. The Augmentative and Alternative Communications Centers at the University of Nebraska are the clinical research partners in this effort.

Sponsors:

This project is sponsored by the National Institutes of Health under a Small Business Innovative Research (SBIR) award to InvoTek. Related previous work has been funded by the Whitaker Foundation.

Selected recent publications in other areas -

Work in collaboration with the Wireless and Video Communications Laboratory at MSU

Hayder Radha and J.R. Deller, Jr., “Multimedia transcaling for the wireless internet,” Proc. IEEE Circuits and Systems Workshop on Wireless Communications, Univ. Notre Dame, Aug. 2001.