Integrated Social Wireless Sensor Networks 

Single-Trial Signal Processing for P300 Based Brain Computer Interface 

 

---

Speaker Recognition and Diarization using Shifted MFCC and GMM 

---

 

Abstract:

 

Investigators: Dr. Ravi Sankar (Electrical Engineering), Rishiraj Mukherjee, Ph.D. (EE), and Tanmoy Islam, Ph.D. Candidate (EE)

 

Funding Information: USF University College: A eLearning and eTeaching Ph.D. Pilot Initiative, 1/1/12-12/31/15

 

Publications:

 

---

Constrained Motion Particle Swarm Optimization for Nonlinear Time Series Prediction Using SVR 

---

 

Abstract:

 

Investigators: Dr. Ravi Sankar (Electrical Engineering), Nick Sapankevych, Ph.D. Candidate (EE)

 

Funding Information: N/A

 

Publications:

 

---

Analysis and Extraction of Music Features  

---

 

Abstract:

 

Investigators: Dr. Ravi Sankar (Electrical Engineering), Steven Brown, Ph.D. (EE)

 

Funding Information: N/A

 

Publications:

 

---

Automatic Modulation Recognition for Cognitive Software Radios 

---

 

Abstract:

 

Investigators: Dr. Ravi Sankar (Electrical Engineering), Sylwester Sobolewski, Ph.D. (EE)

 

Funding Information: N/A

 

Publications:

 

---

VoIP in 4th Generation Cellular Communications

---

 

Abstract:

 

Investigators: Dr. Ravi Sankar (Electrical Engineering), An Le, Ph.D. Candidate (EE)

 

Funding Information: N/A

 

Publications:

 

Sensor Networks, WLAN/WPAN Vertical Handoff 

Abstract: Wireless Sensor Networks (WSN) are deployed in a variety of widespread applications: industrial control and monitoring, home automation, military applications, supply chain tracking, and environmental sensing. Data rate and latency requirements vary depending on the specific application. The USF WSN research group focuses on optimizing sensor network design while considering scalability, power, latency and data throughput constraints. Research also involves cross-layer designing in order to improve sensor efficiency, and exploring the effect of routing protocols on energy efficiency. In addition to sensor networks, this group also focuses on communication and handoff between WLAN and WPAN network technologies. With the prevalence of disparate networking technologies in the consumer and business marketplaces, users will frequently travel between locations that provided coverage of one network service, but not another. This research group is also interested in issues related to heterogeneous, or vertical, handoff so that mobile users of voice, video, or data services may maintain their sessions, possibly under degraded conditions, in a seamless, uninterrupted manner. The specific focus is in developing intelligent protocols that automatically detect and register amongst multiple networks in an efficient manner that does consume unnecessary network resources.

Investigators: Dr. Ravi Sankar (Electrical Engineering), Bill Philips, Ph.D. Candidate (EE)

Funding Information: N/A

Publications:

Robust Speech/Speaker Recognition Systems 

Abstract: Speaker or voice recognition is the task of automatically recognizing people from their speech signals. This allows the use of uttered speech to verify the speaker’s identity and control access to secure services, surveillance, counter-terrorism and homeland security department can collect voice data from telephone conversation without having to access to any other biometric dataset. For biometric security applications, the confidence level of authentication needs to be high. Other applicable areas includes online transactions, database access services, information services, security control for confidential information areas, remote access to computers etc. Speaker recognition history dates back to some four decades, and yet it has not been reliable enough to be considered as a standalone security system. This research project focuses on the enhancement of speaker recognition through fusion of likelihood scores generated by Arithmetic Harmonic Sphericity (AHS) and Hidden Markov Model (HMM) techniques. Our fusion method uses a linear weighted fusion, where the weights are derived from the means of the score distributions. The current research investigation is to enhance speaker recognition including various strategies and techniques to make the system more robust and achieve higher level of performance. Further work will include extension of our algorithm based on the study on speaker and accent modeling techniques. Experiments are being conducted on feature, model and score-level fusion methodologies.

Investigators: Dr.Ravi Sankar (Electrical Engineering), Tanmoy Islam, Ph.D. Candidate (EE), Srikanth Mangayyagari, M.S. (EE)

Funding Information: Florida High Tech Corridor Program, 06/05-08/07

Publications:

Abstract: Speaker or voice recognition is the task of automatically recognizing people from their speech signals. This technique makes it possible to use uttered speech to verify the speaker’s identity and control access to secured services. Surveillance, counter-terrorism and homeland security department can collect voice data from telephone conversation without having to access to any other biometric dataset. In this type of scenario it would be beneficial if the confidence level of authentication is high. Other applicable areas include online transactions, database access services, information services, security control for confidential information areas, and remote access to computers.

Speaker recognition systems, even though they have been around for four decades, have not been widely considered as standalone systems for biometric security because of their unacceptably low performance, i.e., high false acceptance and true rejection. This thesis focuses on the enhancement of speaker recognition through a combination of intra-modal fusion and accent modeling. Initial enhancement of speaker recognition was achieved through intra-modal hybrid fusion (HF) of likelihood scores generated by arithmetic harmonic sphericity (AHS) and hidden Markov model (HMM) techniques. Due to the contrastive nature of AHS and HMM, we have observed a significant performance improvement of 22% , 6% and 23% true acceptance rate (TAR) at 5% false acceptance rate (FAR), when this fusion technique was evaluated on three different datasets – YOHO, USF multi-modal biometric and Speech Accent Archive (SAA), respectively. Performance enhancement has been achieved on both the datasets; however performance on YOHO was comparatively higher than that on USF dataset, owing to the fact that USF dataset is a noisy outdoor dataset whereas YOHO is an indoor dataset.

In order to further increase the speaker recognition rate at lower FARs, we combined accent information from an accent classification (AC) system with our earlier HF system. Also, in homeland security applications, speaker accent will play a critical role in the evaluation of biometric systems since users will be international in nature. So incorporating accent information into the speaker recognition/verification system is a key component that our study focused on. The proposed system achieved further performance improvements of 17% and 15% TAR at an FAR of 3% when evaluated on SAA and USF multi-modal biometric datasets. The accent incorporation method and the hybrid fusion techniques discussed in this work can also be applied to any other speaker recognition systems.

Investigators: Dr. Ravi Sankar (Electrical Engineering), Tanmoy Islam, Ph.D. Candidate (EE), Srikanth Mangayyagari, M.S. (EE)

Funding Information: N/A

Publications:

Integrated DSP/FPGA Lab for Software Defined Radio 

Abstract: The primary objective of this project is to establish a state-of-the-art laboratory for the development of real-time digital signal processing (DSP) systems from algorithm to hardware using DSP, FPGA and hybrid DSP/FPGA rapid prototyping platforms. Several structured laboratory exercises such as sampling, convolution, spectral analysis (FFT), filtering (FIR/IIR) followed by detailed projects on wireless and digital communications, adaptive filters, speech processing, etc. [website: icons.eng.usf.edu/Education_Courses]

Investigators: Dr. Ravi Sankar (Electrical Engineering), Dr. Huseyin Arslan (Electrical Engineering),

John Norstrom, M.S. (EE), Ismail Butun, Ph.D. (EE)

Funding Information: USF Innovative Teaching Grants Program, 04/01/04-
Hardware and Software Donations and/or Discounts from TI, Xilinx, Lyrtech.

Publications:

Multi-modal Biometric Study

Abstract: Human physiological or behavioral characteristics can be used to identify a person. This method of identification is called biometrics. Biometric based recognition systems are often preferred over traditional authentication measures, i.e., password or pins, because unlike passwords or pins we do not lose or forget our own physiological characteristics and it is very resistant to spoofing. There are many different characteristics, also known as modalities, can be used for biometric authentication systems- i.e., fingerprint, face, voice, gait, iris, ear, hand geometry etc. Combination of multiple modalities such as face, voice, and fingerprint can also be used to enhance the performance of the recognition system. USF biometric research group is focused on various aspects to the design of biometric systems: biometric dataset, sensors to gather the data, algorithms to extract features from the data and to compute similarities among the extracted features so that a decision can be made. Biometric research group at USF has already created a multi-modal dataset with variations such as time, location (Indoor or Outdoor), pose etc. Current research includes novel algorithms to extract features from clean (indoor) and noisy (outdoor) data, integration (fusion) of multi-modal biometrics into the identification systems, authentication from distance and crowded areas.
Segmenting different individuals in a group meeting and their speech is an important first step for various tasks such as meeting transcription, automatic camera panning, multimedia retrieval and monologue detection. In this effort, given a meeting room video, we attempt to segment individual person's speech and localize them in the video, based on data from a single audio and video source. The segmentation method is driven by audio and enhanced by video cues. In this project, Bayesian Information Criterion (BIC) was used to segment the feature vector streams and graph spectral partitioning to cluster them. We compare our results with audio based segmentation method and our localization technique with the commonly used mutual information.

ICE-T (E-mammography) 

Abstract: The primary objective of this project is to establish an Interdisciplinary Center of Excellence in Telemedicine (ICE-T) – a bioengineering center for cancer prevention that combines computer-assisted diagnosis, signal processing, telecommunication networking, and statistical data analysis. The ICE-T pilot program will target the study of breast cancer detection, diagnosis and treatment through E-mammography, in general, telemedicine. It is very important to provide mammography services to women in underserved areas via E-mammography, using Internet and wireless networks. With the explosive development of information technologies (Internet, multimedia delivery, data mining), telecommunications infrastructures (wireline, wireless, satellite networks), information processing techniques (imaging, automated detection and decision making), significant advances in telehealth technologies and applications are on the horizon. [website: www.ice-t.eng.usf.edu]

Investigators: Dr. Ravi Sankar (Electrical Engineering), Dr. Wei Qian (Moffitt), Dr. Ji-Hyun Lee (Moffitt), Ying Zhang, Ph.D. Candidate (EE), Kun Li, Ph.D. Candidate (EE)

Funding Information: USF Interdisciplinary Research Development Award, 08/01/04-12/31/06

Publications:

Optimal Data Distribution and Routing for Large MANET 

Artificial Neural Network (ANN) Single-Layer Perceptron (SLP) Training Algorithms 

Abstract: The goal of this project is to research and evaluate training algorithms for SLP ANNs. Specifically, a fully-connected three node SLP will be used as the ANN architecture for regression applications (not the traditional classification application). Metrics for evaluation are robustness, stability, accuracy (avoiding local minima), computational efficiency, and convergence time.

Investigators: Dr. Ravi Sankar (Electrical Engineering), Ying Zhang, Ph.D. Candidate (EE), Srikanth Mangayyagari, M.S. (EE)

Funding Information: Raytheon, Inc. and Florida High Tech Corridor program, 06/01/06-05/31/07.

Publications:

Vehicle Multi-Occupant Detection and Counting 

Abstract: To conduct research technology alternatives, tradeoff analyses and recommended equipment for purchase to autonomously count the number of passengers(including the driver) visible in a vehicle. The recommended equipment must be versatile to accomodate a variety of vehicles( to include motorcycles, minivans, SUV's and light duty trucks) and robust to perform effectively 24/7 in a wide variety of environmental conditions. Technologies to be considered, but not limited to, should include ultrasonic, radar, visible and non-visible spectrum, infrared, thermal, ladar and microwave.

Investigators: Dr. Don Morel (Electrical Engineering), Dr. Ravi Sankar (Electrical Engineering), Dr. Thomas Weller (Electrical Engineering), Dr. Dennis Killinger (Physics)

Funding Information: STS International, Inc., U.S. Army subcontract, 01/05 - 09/05.

Publications:

IGERT Project 

Abstract: This research proposal addresses the design and development of a wireless biosensor network system for home health monitoring in telemedicine. The primary goal is to integrate new and novel biosensors, wireless communication protocols, network system architectures, and technologies to design and characterize an energy-efficient, scalable, and non-invasive biomedical wireless skin sensor network (BWSSN) system. It is envisioned that the proposed prototype system will be customized for new skin based sensors and other off-the-shelf biosensors. The system could become a testing platform for several researchers at USF developing biomedical skin sensor technologies by providing wireless communication capability between the sensor and the associated data processing, monitoring, and diagnostic equipment. It could serve to integrate biomedical sensor research efforts from the physics, chemistry, biology, engineering and medical departments. The combination of sensors, wireless communication, signal and information processing in one platform would be utilized to develop specific disease diagnostic capabilities via continuous, real time analysis of sensor data. [website: www.eng.usf.edu/igert]

Investigators: Dr. Ravi Sankar (Electrical Engineering), Dr. N. Ranganathan (Computer Science & Engineering), Kelvin Rojas, Ph.D. Candidate (EE)

Funding Information: NSF, 08/03-07/08

Publications:

For other projects please click on the images below