The Signal and Speech Research Group (SRG) carries out research on algorithms for signal processing and their application. The group has been led by Professor Michael Savic. Since 1985 the SRG has focused on high-risk, innovative research in all aspects of signal processing. The Current application areas include speaker recognition, language identification, detection of cholesterol deposits in blood vessels, speech recognition, automatic pipeline leak detection, voice character transformation, speaker separation (the cocktail party problem) and others.
SRG has been a pioneer in many areas such as speaker verification, voice character transformation, speaker separation, pipeline leak detection, detection of cholesterol in blood vessels and others. The work of SRG has been nationally and internationally recognized and cited in scientific papers.
The research of SRG has been sponsored by US and foreign research foundations, industry and government agencies.
SRG traditionally
integrates theoretical and experimental studies, and encourages both analytic
understanding and synthetic design capability, such as the development
of fast signal processing computers. The diversity of SRG's research is
valuable, because it offers both breadth and depth in a wide range of signal
processing research.
List of the Projects
Speaker
Verification
Speaker
Recognition I
Speaker
Recognition II
A
Speech Recognition System
Glass
Break Detector
Voice
Character Transformation
Cartoon
Character Voice Simulation
A
Blood Gas Monitor for Divers
Automatic
Pipeline Monitoring
Speaker
Separation
Automatic
Tracking of Acoustic Sources
Automatic
Language Identification
Sonic
Leak Detection System (Phase III)
Detection
of Carotid Arterial Stenosis
Speaker
Verification Using wavelet Features.
Speech
Recognition
Co-Channel
Speaker Separation Phase II
Remote
Cholesterol Plaque Detection
Pilot Projects
Reliability
of Metal Parts
Signal
Recognition
Hardware
Support Projects
Back
to the List of the Projects
Speaker Verification
Verification of the authenticity of a speaker on communication lines. If a speaker claims his identity on the phone, the method can verify the speaker's authenticity. Applications of the system include business transactions, military and forensic applications etc. RPI Proj. 5-24107, Sponsor DOD, 3 years Sept. 1985- Sept. 1988.
Research Staff
Principal Investigators
Prof. Michael Savic (67%)
Prof. John Anderson(33%)
Graduate Students
| Joseph Attili (Ph.D.) | |
| Miles Benson (M.S.) | Daniel Borkowski (M.S.) |
| Keith Carangelo (M.S.) | Amy Carlson (M.S.) |
| Martin Chang (M.S.) | George Chung (M.S) |
| Brian Concoran, (M.S.) | John Delamorte (M.S.) |
| Michael Elser (M.S.) | Charles Chengkwang Houng (M.S.) |
| David Mackersie (M.S.) | Theresa McNeil (M.S.) |
| Theodore Scharer (M.S) | Jay Shain (M.S.) |
| William Shearer (M.S.) | Piush Sinha, (M.S.) |
| Mary Vetter (M.S.) | John Voutsas (M.S.) |
Publications
M. I. Savic and J. B. Attili "A TMS320 Based Speaker Verification System, "Speech Research Symposium IV, at Princeton, New Jersey, June 10-11 (1985).
M. I. Savic and J. B. Attili, "A New Speaker Verification Method," Speech Research Symposium V, at Baltimore, Maryland, February 18-19 (1987).
J. Attili, M. I. Saviic and Daniel G. Borkowski, "Speaker Dependent Features for Text-Independent Verification," Proceedings of the IEEE Carnahan Conference on Security Technology, pp. 59-63, July 1987.
J. Attili, M. I. Savic and J. P. Campbell, Jr., "A TMS32020-Based Real Time, Text-Independent, Automatic Speaker Verification System," Proc. ICASP 88, Vol. I, pp. 599-602, Apr. 1988.
J. B. Attili, J. B. Anderson and M. I. Savic "On the Estimation Error Associated with the Autocorrelation Method LPC Estimate," Proc. IEEE International Conference on Communications (ICC 88), Vol. 3, pp. 1608-1613, Philadelphia, PA June 1988.
M. I. Savic, "Speaker Verification Using the Hidden Markov Model", Proc. IEEE Confer. in Utica, May 91.
M. I. Savic and Jeffrey Sorensen, "Speaker Verification Using Segmentation into Phonemes", SRS-XI, Boston, June 1991.
M. I. Savic and J. Sorensen, "Phoneme Based Speaker Verification", Proc. ICASSP92.
J. S. Sorensen and M. I. Savic, "Hierarchical Pattern Classification
for High Performance Text-Independent Speaker Verification Systems", Proc.
SRSXIII, Baltimore MD, pp. 49-52 June 1993.
Back
to the List of the Projects
Speaker
Recognition I
Automatic identification of a speaker on communication lines. RPI Proj. 5-24111; Sponsor private industry. Two years March 1989 -- March 1991
Research Staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Sunil Gupta (Ph.D.) | |
| Willie Baldwin (M.S.) | Mahesh Chugani (M.S.) |
M. I. Savic and S. K. Gupta, "Adaptive Vocal Tract Model Emulates the Vocal Tract of the Speaker," Proc. ISMM, Dec. 1988.
M. I. Savic, S. K. Gupta and J. B. Attili, "RPI Speaker Verification System-A Fresh Look," Proc. SRS-VII, June 1988.
M. I. Savic and S. K. Gupta, "Adaptive Multi-Feature Based Speaker Verification System". Proc. of Military and Government Speech Tech `89, pp. 243-247, Oct. 1989.
M. I. Savic and Jeffrey Sorensen, "Improved Speaker Verification by Phonetic Segmentation," Proc. Speech Research Symposium X , Oct. 1990.
S. K. Gupta and M. I. Savic, "Text-Independent Speaker Verification Based on Broad Phonetic Segmentation of Speech," Digital Signal Processing Journal, pp. 69-79, Vol. 2, No. 2, Academic Press, April 1992.
M. I. Savic and S. K. Gupta, "An Adaptive Multifeature Based Speaker Verification System, "International Journal for Mini and Micro Computers, pp. 1-8, Vol. 12, No. 1, 1990.
M. I. Savic, S. K. Gupta and J. Sorensen, "The RPI Speaker Verification System," Proc. of the Speech Research Symposium IX, (1989),.
M. I. Savic and S. K. Gupta, "Variable Parameter Speaker Verification System Based Hidden Markov Modeling," Proc. ICASSP 90, pp. 281-284, Apr. 1990.
M. I. Savic and S. K. Gupta, "On the Use of HMM Based Constrained Clustering for Speaker Verification," Proc. Speech Tech 90, pp. 128-132, Apr. 1990.
M. I. Savic, "Speaker Verification Using the Hidden Markov Model," Proc. Conference Record, IEEE Confer. in Utica, May 91.
S. Gupta and M. Savic, "Speaker Verification
Using Neural Networks," Proc. SPEECHTECH 91, Apr. 91.
Back
to the List of the Projects
Speaker
Recognition II
Automatic system for Speaker Identification by his voice. RPI Proj. 5-25676. Sponsor private industry. Two years Sept. 1988-Sept. 1990.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Nimal Ratnayake (Ph.D.) | Jeffrey Sorensen (Ph.D.) |
Publications
M. I. Savic and Jeffrey Sorensen, "Speaker Verification Using Segmentation into Phonemes," SRS-XI, Boston, June 1991.
M. I. Savic and J. Sorensen "Phoneme Based Speaker Verification," Proc. ICASSP92, Vol. II, pp. 165-168, San Francisco, March 1992.
M. I. Savic and Jeffrey Sorensen "Automatic Speaker Verification," SRSXIII. June 1993.
J. S. Sorensen and M. I. Savic, "Hierarchical Pattern Classification for High Performance Text-Independent Speaker Verification Systems", Proc. SRSXIII, Baltimore MD, pp. 49-52 June 1993.
J. S. Sorensen and M. l Savic, "Hierarchical
Pattern Classification for High Performance Text-Independent Verification
Systems", ICASSP94.
Back
to the List of the Projects
A Speech Recognition System
Automatic recognition of the words of a speaker. Applications include giving verbal instructions to a device such as a typewriter, airplane, car or other. RPI Proj. 5-25690. Sponsor private industry. Two years, Sept. 1988-Sept. 1990.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Nimal Ratnayake (Ph.D.) | |
| Igal Brodetsky (M.S.) | Carl Tung (M.S.) |
Publications
N. Ratnayake, M. I. Savic and J. Sorensen "Use of Semi-Markov Models
for Speaker-Independent Recognition of Phones", Proc. ICASSP92, vol. 1
pp. 565-568.
Back
to the List of the Projects
Glass
Break Detector
The Glass Break Detector is a device that recognizes nonparametric signals.
It can be trained to recognize with high accuracy the sound of breaking
glass, while ignoring all other sounds. RPI Project 5-25709; Sponsor private
industry, 2 years, Feb. 1988 - Feb. 1990.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Sew Hwee Tan (Ph.D.) | Willie J. Baldvin (M.S.) |
| George C. Chung (M.S.) | Robert Crovella (M.S.) |
Publications
M. I. Savic and S. Tan, "A TMS32020 Based Signal Recognition System with a Neural Network Classifier," Proc. ISMM, Dec. 1988.
M. I. Savic "A New Class of Neural Networks," Proc. SRS-VIII (invited), Stillwater Oklahoma, March 15, 1989.
S. Tan and M. I. Savic, "A Training Algorithm for the Sequential Classification of Patterns Using Static Nets," Proc. ICASSP 91, pp. 3421-3424, May 91.
M. I. Savic and Sew Hwee Tan, "Self-Sizing, Non-Linear Neural Nets," Proc. SRS-XI, Boston, June 1991.
M. I. Savic and Seow-Hwee Tan, "A New Class of Neural Networks Suitable
for Intelligent Control," Proc. IEEE International Symposium on Intelligent
Control 1989, pp. 418-423, Sept. 1989.
Back
to the List of the Projects
Voice Character Transformation
Transformation of one person's voice into another person's voice. Applications in the entertainment industry, for speaker adaptation in communications and other. Using speaker adaptation a transmitter or receiver can be optimized for a particular (normalized) speaker. Other speaker's voices can than be changed into the voice of the normalized speaker. RPI Project 5-25961, Sponsor Walt Disney Inc. 15 months June 1988-Sept. 1989.
Research Staff
Principle Investigator
Prof. Michael Savic
Graduate Students
| Il-Hyun Nam, (Ph.D.) | Kevin Flemming (M.S.) |
Back
to the List of the Projects
Cartoon Character Voice Simulation
Simulation of voices of cartoon characters. Sponsor Walt Disney Inc. RPI Project 5-25962, 1 year Nov. 1989-Nov. 1990.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Il-Hyun Nam, (Ph.D.) | John McDonough (M.S.) |
Publications
M. I. Savic and Il-Hyun Nam, "Voice Personality Transformation" Digital Signal Processing Journal, pp. 107-110, Vol.1 No. 2, 1991.
M. I. Savic and Il-Hyun Nam, "A System for Voice Personality Transformation," Proc. Speech Tech 90, pp. 118-122, Apr. 1990.
M. I. Savic and Il-Hyun Nam, "Neural Network Based Voice Personality Transformation," Proc. MICRO 90, pp. 135-137, Apr. 1990.
M. I. Savic and Il-Hyun Nam, "The RPI Voice Transformation System," Proc. Speech Research Symposium X, (invited), Oct. 1990.
M. I. Savic, "Applications of the TMS320 in Voice and Signal Processing,"
Proc. TI TMS320 Conference, pp. 17-38, July 31, 1991.
Back
to the List of the Projects
A Blood Gas Monitor for Divers
If divers emerge too fast to the surface of water, bubbles of oxygen in their blood can cause instant death. The monitoring system warns the diver when the amount of bubbles becomes critical. RPI Proj. 5-25533, 4 months, June 1990-Sept.1990 Sponsor Walt Disney Inc.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Student
Sew Hwee Tan (Ph.D.)
Back
to the List of the Projects
Automatic Pipeline Monitoring
Development of a system for monitoring leaks in pipelines that carry explosive gasses under pressure. The purpose of the system is to give an automatic warning if there is a leak in underground pipelines that carry gas under pressure. A leak in these pipelines can generate a highly explosive cloud which when ignited can cause extensive material and human losses. RPI Proj. 5-27030, three years, Sponsor Texaco. Sept. 1990-Dec. 1994.
Research staff
Principal Investigator
Prof. Michael Savic
Research Associate
Dr. Song Wu
Graduate Students
| Igal Brodetsky (Ph.D.) | Nimal Ratnayake (Ph.D.) |
| Chi Yao (M. S.) |
Publications
M. I. Savic "Pipeline Leak Detection," PACC 1990, Apr. 90, Houston Texas.
M. I. Savic and I. Brodetsky, "Pipeline Leak Detection," PACC 1991, Apr. 91, Houston Texas.
Brodetsky and M. I. Savic "Leak Monitoring System for Gas Pipelines,"
Proc. ICASSP93, pp. III -17 to III - 20, Minneapolis, April 27-30, 1993.
M. I. Savic, "Acoustic Gas Leak Monitoring", Proc. 1995 International
Gas Research Conference, pp. 304-313, Cannes, France (Nov. 1995).
Back
to the List of the Projects
Speaker Separation
Extraction of a single speaker from an environment comprising a number of interfering speech and non-speech signals. If many speakers speak at the same time the system can separate one speaker at the time and minimize the interference. Applications include hearing aids, flight recorders, elimination of interference in communications and other. RPI. Proj. 5-25600, five years, Oct. 1991-Apr. 1996.
Research Staff
Principal Investigator
Prof. Michael Savic
Visiting Professor
Prof. Nurgun Erdol
Graduate Students
| Vincent Amuso (Ph.D.) | Dan Benincassa (Ph.D) |
| Huiqin Gao(Ph.D.) | Jeffrey Sorensen, (Ph.D.) |
| John Alvermann (M. S.) |
Publications
J. Sorensen, M. Savic, H Gao and V. Amuso, "Speaker Separation" Proc. SRSXII, Rutgers University, NJ. 1992.
J. S. Sorensen, M. Savic, H. Gao and V. J. Amuso, "A Tracking Multi speaker Model or Speech Separation Applications," Proc. SRS-XII, Piscataway, New Jersey, June1992.
J. Sorensen, M. I. Savic and H. Gao "A deconvolution Based Algorithm for Co-channel Speaker Separation", Proc. IEEE Dual Use Conference, Utica, May 1993.
M. Savic, H. Gao and J. Sorensen "Co-channel Speaker Separation Based on Maximum-Likelihood Deconvolution, ICASSP 94.
H. Gao, M. I. Savic and J. Sorensen, "Minimum Variance Deconvolution Based-Speech Enhancement System for a New Generation of Hearing Aids", Proc. 20th Northeast Bioengineering Conference, IEEE, Springfied, MA, pp. 71-74, (March 1994).
M. I. Savic, V. Amuso and R. Benincassa, "Speaker Separation" Proc. IEEE Dual Use Conference, Utica NY, (May 1995).
D. Benincassa and M. I. Savic "Co-channel Speaker Separation Using Constrained
NonlinearOptimization" Proc. IEEE ICASSP97 pp. 1195 - 1199, Munich
Germany (1997).
Back
to the List of the Projects
Automatic Tracking of Acoustic Sources
The project involves the determination of the geometric location of the speaker based on his speech. The environment is a room or the stage. RPI Proj. 5-25661, sponsor Walt Disney Inc. two years, Dec. 1991- Dec. 1993.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Etienne Marceret (Ph.D.) | Xu Xiang (M. S.) |
Back
to the List of the Projects
Automatic Language Identification
Automatic identification of languages spoken by different speakers. The method can automatically determine the language of an utterance, for example Chinese, English, Korean, etc. Applications include devices for automatic verbal response in the language of the caller. RPI. Proj. 5-24003, sponsor DOD, three years, Oct. 1992-Oct. 1995.
Research staff
Principal Investigator
Prof. Michael Savic
Visitng Professor
Prof. Nurgun Erdol
Graduate Students
| Huiquin Gao (Ph.D.) | Shahid Mahmood (Ph.D.) |
| Etienne Marceret (Ph.D.) | Elena Acosta (M.S.) |
| David Cameron (M.S.) | Ghandi(M. S.) |
| James Heck (M.S.) | Doris Lin (M.S.) |
| Theresa Lorenzo (M.S.) | Larry Mianis (M.S.) |
| Reggie Mijjins (M.S.) | Gandhimath Vaithilingam (M.S.) |
M. I. Savic, E. Acosta, and S. K. Gupta, "Automatic Language Identification," Proc. ICASSP 91, pp. 817-820, May 91.
M. I. Savic and Etienne Marceret, "Automatic Language Identification," Proc. SRS XII, Rutgers University, New Jersey 1992.
E. Marcheret, M. I. Savic and N. Ratnayake ,"Using The Hidden Semi-Markov Model for Language Identification". SRSXIII, Baltimore MD, June 1993.
E. Marcerett and M. I. Savic "The Rensselaer Automatic Language Identification System", Proc. SRSXV, Baltimore, MD June 1995.
E. Marcheret and M. I. Savic "Random
Walk Theory Applied to Language Identification," Proc. IEEE ICASSP97
pp. 1119 - 1123, Munich Germany (1997).
Back
to the List of the Projects
Sonic Leak Detection System (Phase III)
Development of a system for recognition of the sound of a leak from underground pipelines that carry explosive gasses (such as Ethylene) under pressure. The purpose of the system is to automatically shut off the pump, give a warning signal and indicate the location of the leak. A leak can generate a highly explosive cloud which when ignited can cause extensive material and human losses. RPI Proj. 5-27030, three years, Sponsor Texaco, Sept. 1990-Nov. 1993.
Research staff
Principal Investigator
Prof. Michael Savic
Research Associate
Dr. Song Wu
Graduate Students
| Seetha Kodiyalam(Ph.D.) | Chi-Yao Young (M.S) |
Back
to the List of the Projects
Detection of Carotid Arterial Stenosis
A new method for the "Detection of Carotid Arterial Stenosis" that is a cerebrovascular disease caused by cholesterol deposits inside the artery. Our method is based on digital signal processing and it was tested on 35 patients.
Research staff
Principal Investigator
Prof. Michael Savic
Visiting Scientist
Zlatko Macek M.D.
Graduate Students & Undergraduate Students
| Mahesh Chugani (Ph.D.) | Joe Chen (M.S.) |
| James Heck (M.S.) | Richard Kim (M.S.) |
| John Moisan (M.S.) | Kalpesh Pankaj Parekh (M.S.) |
| Minghong Tang (M.S.) | Anh Do |
| James Hatch | John Yun |
Publications
M. I. Savic, M. Chugani, Z. Macek, M. Tang and A.Husain, "Detection of Stenotic Plaque in Arteries", Proc. EMBS, vol , pp. 317-318 (Nov. 1993).
M. Chugani and Michael Savic, "A Mathematical Model of Blood Flow through Stenotic Arteries", Proc. of the 20th Annual Northeast Bioengineering Conference, pp. 13-19 (March 1994).
M. I. Savic et all, "Detection of Cholesterol Plaque in Arteries using
Digital
Signal Processing", Proc. IEEE Dual Use Conference, IEEE. Utica
NY, pp. 465 - 472 (May 1994).
M. I. Savic et all, "Detection of Cholesterol Plaque in Arteries using
Digital Signal Processing", Proc. IEEE ICASSP95, Detroit Michigan
( 1995)
Back
to the List of the Projects
Speaker Verification Using Wavelet Features
The purpose of this project is to investigate the suitability of wavelet features for speaker verification. This project will lead to a better model of the vocal tract and it will most likely improve the performance of many other systems such as Speech recognition, Speech synthesis and other.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Nabil Hijazi (Ph.D.) | Dennis Velasco (M.S.) |
Publications
N. Hijazi and M. I. Savic "Speaker Verification Using Wavelet Features",
Proc. IEEE Dual Use Conference,. Utica NY, (May 1995).
Back
to the List of the Projects
Speech Recognition
A speech recognition system based on the recognition of phonemes.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Student
Nimal Ratnayake (Ph.D.)
Publications
N. Ratnayake, M. I. Savic and J. Sorensen "Use of Semi-Markov Model's
for Speaker-Independent Recognition of Phones," Proc. ICASSP92, Vol. I,
p.p. 565-568, San Francisco, March 1992.
Back
to the List of the Projects
Co-Channel Speaker Separation Phase II
Extraction of a single speaker from an environment comprising a number of interfering speech and non-speech signals. If many speakers speak at the same time the system can separate one speaker at the time and minimize the interference. Applications include flight recorders to find out who said what before a crash, in communications to eliminate interference from other speakers, etc. RPI. Proj. A-90074Y, One year, Avg. 1996-July. 1997.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Vincent Amuso (Ph.D.) | Dan Benincassa (Ph.D) |
Back
to the List of the Projects
Remote Cholesterol Plaque Detection
The objective of this research is to develop a simple and inexpensive system for remote detection of plaque in arteries. This research will lead to a system affordable to physicians and health centers with small budgets. Consequently, this will provide the means for testing and follow up of a large segment of the population. In addition, the examination will be without risk and with minimum discomfort to the patient. RPI project #70552, Sponsor Sunward Electronics, 18 months.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Hyeok-Jae Chi (M.S.) | Illya Gerdes (M.S.) |
| Sanghun Sin (M.S.) |
Back
to the List of the Projects
Pilot Projects
Reliability of Metal Parts
A system is being developed that will indicate when metal parts become unreliable and are about to break.
Research staff
Principal Investigators
| Prof. Michael Savic ECSE | Prof. Aleksander Ostrogorsky MECH. ENG. |
Graduate Students
| Kenneth Lyons (M. S.) ECSE | John Ryan (M. S.) MECH. ENG. |
Back
to the List of the Projects
Signal Recognition
New methods for signal recognition are being developed. These methods can be applied to speaker verification, pipeline leak detection and other.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Sandra Kwei (M. S.) | Greg Mesolella (M. S.) |
| Anila Razaq(M. S.) |
Back
to the List of the Projects
Hardware Support Projects
The Signal and Speech Research Group (SRG) traditionally integrates theoretical and experimental studies, and encourages both analytic understanding and synthetic design capability, such as the development and design of high speed microcomputers for signal processing. Most of SRG's projects have been implemented on these computers. Hardware development has been supported by manufacturers of fast signal processors. Rensselaer Signal and Speech Research Laboratory includes the well-equipped TMS320 DSP Laboratory, which was installed with the help of Texas Instruments Incorporated. This laboratory has been certified by Texas Instruments to promote excellence in engineering. Many hardware and software projects based on TMS320 fast signal processors have taken place in the TMS320 laboratory.
Research staff
Principal Investigator
Prof. Michael Savic
Graduate Students
| Igal Brodetsky (Ph.D.) | Nabil A. Hijazi (Ph.D.) |
| Nimal Ratnayake (Ph.D.) | |
| Martin Chung (M.S.) | John Delamorte (M.S.) |
| Kalpesh Parekh (M.S.) | Carl Tung (M.S.) |
Publications
M. I. Savic "Implementations of the TMS320," TI conference, Houston 1991.
M. I. Savic, Nabil A. Hijazi, Igal Brodetsky, Kalpesh Parekh, Nimal Ratnayake and Carl Tung "Design and Implementation of a Parallel Signal Processing Computer Based on the TMS320C3x Processor," TI conference, Houston 1992.