Wireless Networking Projects

 

Networking Projects

 

Digital Signal Processing Projects

• G. Edwards, Advanced Handoff Algorithmsfor Microcellular Communication using Fuzzy Techniques, September 1996.

   

• G. Edwards and R. Sankar, Microcellular Handoff Using Fuzzy Techniques, Wireless Networks, August 1997, (In second revision).

   

• G. Edwards and R. Sankar, A Predictive Fuzzy Algorithm for High Performance Microcellular Handoff, IEEE Global Telecommunications Conference (Globecom), Phoenix, AZ, November 1997, (accepted).

   

• G. Edwards, A. Kandel, and R. Sankar, Fuzzy Control for Microcellular Handoff, International Journal of Fuzzy Sets and Systems, June 1996, In Review.

   

• G. Edwards and R. Sankar, Fuzzy Control for Microcellular Corner Effect, Proc. Fifth IEEE International Conf. on Fuzzy Systems (FUZZ-IEEE '96), New Orleans, LA, September 1996, pp. 1912-1916.

   

• G. Edwards and R. Sankar, Handoff using Fuzzy Logic, Proc. IEEE Globecom '95, Singapore, November 1995, pp. 524-528.

   

• G. Edwards and R. Sankar, Fuzzy Control for Microcellular Handoff, Proc. JAMCON '95 Communications Conference , Jamaica, August 1995, pp. 68-73.

   

• V. Kapoor, G. Edwards, and R. Sankar, Handoff Criteria for Personal Communication Networks, Proc. IEEE International Conf. on Communications (ICC), New Orleans, LA, May 1994, pp. 1297-1301.

   

• G. Edwards and R. Sankar, A New Handoff Algorithm Using Fuzzy Logic, Proc. IEEE Southeastcon '94, Miami, FL, April 1994, pp. 89-92.

   

• V. Kapoor, An Adaptive Prediction Based Handoff Algorithm and its Simulation for Microcellular CDMA, November 1993. July 1992.

   

• R. Sankar and L. Civil, Traffic Monitoring and Congestion Prediction Using Handoffs in Wireless Cellular Communications, IEEE 47th Annual International Vehicular Technology Conference (IEEE VTC), Phoenix, AZ, May 1997, pp.520-524.

   

• R. Sankar and L. Civil, Intelligent Traffic Monitoring System Using Wireless Cellular Communications, IEEE Southeastcon '97, Blacksburg, VA, April 1997, pp. 210-214.

   

• R. Sankar and L. Civil, Traffic Monitoring and Congestion Management using Cellular Communications Technology, USF Research Council, June 1996, (Final Report, 80 pages).

   

 

---

Project Description - Network Communications

---

 

VBR Video Traffic Modeling

 

The characteristics of the VBR video traffic vary with the coding algorithm and the image content. We think that the selection of dynamic bandwidth allocation methods depends on the customer’s service requirements and the source characteristics. In this paper we propose a bandwidth allocation method for the stationary VBR traffic based on prediction. Our study concentrates on the following: (1) How to determine the allocated bandwidth and buffer space according to the requirements on the loss rate and utilization. (2) How to allocate the bandwidth and the buffer space accordance with the loss rate and delay upper limit. (3) Analyzing under what condition, the requirements of loss rate, delay upper limit and utilization can be satisfied simultaneously. No knowledge about the source characteristics is required in advance for our model and all the control parameters can be obtained by measuring the mean of input and the prediction error on-line.

 

Linear prediction methods have been widely used to forecast the bit rate varying of the coded video stream. In this research we show that: (1) generally, linear predictors are only suitable to predict the bit rate variation of the stream not containing scene changes or during a scene, (2) a proposed method based on scene change identification can improve the forecast performance, and (3) a dynamic bandwidth allocation scheme based on a smoothing algorithm can guarantee the delay upper limit not larger than a group frame cycle and make full use of the allocated bandwidth.

 

Dynamic Bandwidth Allocation for VBR Video Traffic

The characteristics of the VBR video traffic vary with the coding algorithm and the image content. The selection of dynamic bandwidth allocation methods critically depends on the customer's service requirements and the source characteristics. We have proposed a method to determine bandwidth allocation and buffer space according to the loss rate and utilization requirements for the stationary VBR traffic based on prediction. Linear predictors fail when there are lot of scene changes (non stationary traffic). We have proposed a way to improve the forecast performance by scene change identification and traffic smoothing.

 

---

Traffic Management and QoS Issues: Broadband (ATM) Networks

 

Research Objectives

The real-time video traffic is expected to occupy a large bandwidth of broadband networks. Dynamic bandwidth allocation schemes are necessary in order to efficiently utilize the network resources (e.g., bandwidth, buffers) and maximize the number of video sessions that can be supported with existing resources. Proper online adaptation to the changing bandwidth requirements of the highly correlated video traffic at regular intervals is necessary to achieve the desired queue performance. This issue of reducing the complexity and yet providing better response time for effectively supporting real-time traffic in high-speed ATM networks, is one of our key research goals.

The Advance reservations for the resources (bandwidth) has been an active area of research. Most of the current techniques may be categorized as semi-static over a certain period of time until further renegotiation for bandwidth takes place. The dynamic adaptability with the changing bandwidth requirements based on online traffic measurements is open for investigation. Even with the semi-static resource allocations, once the bandwidth is reserved for a session, the utilization of the bandwidth may be of concern if the reserved bandwidth is not fully utilized. This is where we need short-term controlling mechanisms for bandwidth allocation in order to maximize the utilization. The cell-level scheduling facilitates such short-term controlling mechanisms. Frequent renegotiations may not be possible in an ATM network environment where bandwidth requirements for both ongoing sessions and sessions that are in call-admission phase are changing. The online traffic measurement-based adaptive bandwidth allocation schemes are essential for a quick adaptation to the changing traffic rates and minimizing the number of renegotiations for bandwidth, especially in the context of highly correlated VBR video traffic. The linear prediction of video traffic is being investigated using autoregressive (AR) models. It is usually assumed that the nodes (ATM switches) will reserve the bandwidth as requested by predictors. This may not be the situation when we multiplex a number of video sessions sharing the bandwidth (statistical multiplexing gain). The other issue that arises from prediction of video traffic is the effect of error in traffic estimation on the queues. Our video traffic management architecture addresses these issues.

 

Predictor-based Architecture for an Adaptive Traffic Management of VBR Video

The correlation properties of video traffic make predictor-based bandwidth allocation schemes attractive. A novel predictor-based architecture that addresses the above issues through a dynamic bandwidth allocation at burst level (milliseconds) and a short-term resource management through cell-scheduling (micro secs) for video traffic has been proposed and analyzed. In order to reduce computational requirements, linear predictors (regressive) of VBR video traffic were employed and to minimize the ill-effects of traffic prediction errors on queueing system, a short-term controller was used to provide a quick reactive control (overestimated bandwidth of some of the sessions is shared among the underestimated sessions). It has been shown that dynamic adaptive techniques outperform static allocation schemes in terms of queue lengths and delays. We propose an integrated framework for VBR video traffic management based on traffic prediction that facilitates the online adaptation to the changing traffic rates as shown in Figure 1.

 

 
Figure 1: The integrated framework

 

The correlation properties of the VBR video traffic make traffic prediction possible and based on the predictor estimates online adaptation to traffic rates can be acheived. Based on the traffic estimates for future adaptation intervals, the predictor system dynamically allocates the bandwidth to various ongoing video sessions. The short-term controller (STC) works at cell-level scheduling while the predictor system works at the burst-level (frame or few tens of milliseconds). The purpose of the STC is to reduce the effects of prediction errors on the queues of individual sessions and also at the same time exploit the statistical multiplexing gain across the sessions. Figure 2 depicts proposed Predictor and Short-term controller (Predictor-STC) architecture. The predictor module (PM) estimates the amount of bandwidth of various sessions, and accordingly reserves the bandwidth per session. The STC schedules the cells according to the reserved bandwidth by the PM, but provides a quick reactive control to the errors in bandwidth estimation of the PM, through cell/slot adjustments of various sessions. Thus STC is very useful as there will be a certain amount of error in the estimation of the long term predictor. The STC plays a crucial role in maximizing the statistical multiplexing gain. The traffic of an underestimated session is served with the minimum bandwidth MinBW, in addition to STC adjustments from the overestimates of other session's bandwidth. This can also be helpful in fine tuning the QoS provided. The STC also acts as a traffic regulator conforming to the peak-rates of various sessions. Thus STC provides a reactive control and is indispensable in the predictor-based server architectures. This approach also gives a closer way of looking at the sessions instead of treating them as en-masse sessions of various classes and provides a mechanism of online adaptation to changing rates of VBR video traffic.

 

Figure 2: The Predictor-STC system

 

The total capacity (bandwidth) that is available for the predictor system for the allocation among the sessions is obtained from the Call-admitter. During the call-admission phase, the Call-admitter relies on prior statistical characterization of the prospective video sessions that may not be accurate enough leading to an approximate estimation of bandwidth requirement. Thus the QoS Change and Bandwidth Renegotiation (QSCBR) module plays an important role for the necessary renegotiations for the desired bandwidth that may be needed by the sessions after admission. The Predictor-STC system provides a mechanism for online adaptation to traffic rates of individual sessions and at the same time exploits the statistical gain across the sessions thereby decreasing the need for frequent bandwidth renegotiations. The change in end-user based picture quality requirements are considered as QoS changes (equivalently translated to the bandwidth requirements) of the ongoing video sessions that are taken care by the QSCBR module. Work is underway to further study the decrease in bandwidth renegotiations that are possible due to the exploitation of statistical gain across the sessions.

 

---

For Details See the Relevant Publications

---

 

• G. Chiruvolu, R. Sankar, and N. Ranganathan, VBR Video Traffic Management using a Predictor-based Architecture Computer Communications, July 1999, (in print).

   

• G. Chiruvolu, R. Sankar, and N. Ranganathan, Adaptive VBR Video Traffic Management for Higher Utilization of ATM Networks , ACM/SIGCOMM Computer Communication Review , Vol. 28, No. 3, pp. 27-40, July 1998.

   

• G. Chiruvolu, R. Sankar, and N. Ranganathan, An Adaptive Scheme for Better Utilization with QoS Constraints for VBR Video Traffic in ATM Networks, Third IEEE Symposium on Computers and Communications, Athens, Greece, June 1998.

   

• G. Chiruvolu, T. Das, R. Sankar, and N. Ranganathan, A Scene-based Generalized Markov Chain Model for VBR Video, IEEE International Conf. on Communications (ICC), Atlanta, GA, June 1998.

   

• G. Chiruvolu, R. Sankar, and N. Ranganathan, Issue and Approaches Towards VBR Video Traffic Management in ATM Networks, IEEE Southeastcon '98, Orlando, FL, April 1998, pp. 306-310.

   

• G. Chiruvolu and R. Sankar, An Approach Towards Transportation and Efficient Resource Management of VBR Video traffic, IEEE International Conference on Communications, pp. 550-554, June 1997.

   

• G. Chiruvolu, K. J. Christensen and R. Sankar, Short-Term Resource Management for Better ATM QoS , Sixth International Workshop on Network Operating System Support for Digital Audio and Video, pp. 31-34, 1996.

 

---

 

Resource/Bandwidth Management

For proliferation of ATM to the desktop, it is critical that legacy LAN internetwork traffic is supported over the public ATM network. Several techniques that provide LAN interconnetivity via ATM include: tunneling, maintaining permanent ATM resources in the ATM network, and acquiring resources on-demand. However, these schemes have their own pros and cons in supporting LAN traffic requiring QoS quarantees.

The primary goal of this research is to develop an integrated approach based on Demand Allocation with Channel Reuse for the transportation of Inter-LAN traffic over ATM.

 

---

For Details See the Relevant Publications

---

 

• S. Varada, R. Sankar, and Y. Yang, ATM to the Desktop in an Existing LAN Environment, The First NDSU Workshop on ATM Networking, Frago, ND, pp. 86-100, Aug. 1996.

   

• S. Varada and R. Sankar, Bandwidth Allocation for Connectionless Traffic in ATM Networks, IEEE Southeastcon '95, Raleigh, NC, pp. 128-132, Mar. 1995.

 

---

 

Congestion Control for ABR Services in ATM Networks

The ATM Forum has adopted the rate-based scheme as the standard for ABR services. Effective ABR services are essential for ATM LAN Emulation. WE have proposed a new algorithm (or scheme) that has equivalent or better performance than existing schemes, but requires only O(1) computation. This O(1) computation is independent of both the number and rates of the connections. The scheme, called the "USF scheme" is compared via simulation to the existing EPRCA, MIT and OSU schemes. Results show that the less complex USF scheme provides better throughput, delay, and fairness than the existing schemes.

 

---

For Details See the Relevant Publications

---

 

• K. T. Ma, Congestion Controls for ABR Services in ATM Networks, Masters Thesis, Univ. of South Florida, Dec. 1996.

   

• K. Ma, R. Sankar, and K. Christensen, A New Explicit Rate-Based Congestion Control Scheme for ABR Services, 22nd IEEE Annual Conference on Local Computer Networks, Oct. 1997.

 

---

 

Gigabit ATM Network Testbed for Traffic Management Research

The efficient transportation of real-time Variable Bit Rate (VBR) video traffic in the high-speed networks has been an area of active research. The VBR video traffic characteristics having heavy tail distribution, high variance and correlation properties are quite complex to be captured by a single traffic model. While many methods have been proposed in the literature focusing on various aspects of the VBR video traffic characteristics and their impact on the traffic management, a wider perspective of various issues involved in the efficient transportation of VBR video traffic with high utilization of network resources is imperative. Moreover one important issue is the simplicity with which the bandwidth allocation and scheduling schemes can be executed online with real-time constraints in the future Gigabit networks. The correlation properties of the VBR video traffic make the predictor-based online traffic adaptation techniques attractive. In order to reduce the effects of the prediction errors on the queueing system, we have designed a novel short-term controller (STC) that works at the cell-level and the system is called the Predictor-STC system. Simple online prediction based bandwidth allocation scheme and a scheduling algorithm for the STC suitable for implementation in High-speed networks have been designed. We are developing an integrated framework addressing the issues in the VBR video traffic management based on the Predictor-STC exploiting the correlation properties.

While most of our current research relies on simulation results, the use of Gigabit Network Kit (GNK) should provide a better insight into how our architecture and other existing methods work vis-a-vis the QoS that can be provided to the end-user. We intend to evaluate the proposed architecture using the Gigabit Network Kit. More details of the research can be found at here

 

---

 

Token-Ring LAN

Simulation modeling and performance analysis of networks including traditional local networks such as token-ring and high-speed networks such as FDDI have been conducted. The token-ring local area network protocol (IEEE 802.5) was simulated using CSIM, a process oriented simulation language based on C. The model behavior was analyzed and its and its performance was compared with an analytical model. The effect of buffer size and service schemes on the network performance was studied.

Analytical and simulation models for studying the effect of integrated traffic (voice and data) on a token-ring network have been developed which can be extended to video traffic as well. The model provides the selection of optimum voice packet length, maximum number of voice stations that the network can support, relationships between voice packet delay and number of voice stations and between voice and data traffic in the same network, and statistical characteristics of voice packet delay. This can be used as practical design rules for implementing multimedia integrated services using existing technology and standards.

A dynamic protocol for token-ring network with unbalanced load conditions was designed by selecting the value of token_holding_time according to the load (or the queue length) at the node. This achieved both fairness in and better performance in terms of reduced waiting delay and buffer occupancy.

 

---

For Details See the Relevant Publications

---

 

• G. Edwards, R. Sankar, and P. Roth, CSIM Simulation of a Token Ring LAN , Transactions of the Society for Computer Simulation, Vol. 9, No.1, pp. 39-58, March 1992.

   

• G. Edwards and R. Sankar, Modeling and Simulation of Networks using CSIM, SIMULATION, Vol. 58, No.2, pp. 131-136, Feb. 1992.

   

• R. Sankar and G. Edwards, Effect of Service Schemes and Buffer Size on Token Ring Network Performance, Proc. ISMM International Symposium MICRO '90, Montreal, Canada, May 1990.

   

• P. Roth and R. Sankar, Token-Ring Network - A Comparative Simulation Study, Proc. 21st Pittsburgh Modeling and Simulation Conference, Pittsburgh, PA, May 1990.

   

• G. Edwards, Modelling and Performance Analysis of Token Ring Network, M.S. Thesis, November 1988.

   

• R. Sankar and J. Murphy, Experiences in Local Networking USF, Proc. 15th IEEE Conf. on Local Computer Networks, Minneapolis, MN, October 1990, pp. 137-142.

   

• W. Li and R. Sankar, Real-Time Voice and Data Integration in Token Ring LAN, Proc. IEEE Southeastcon '93, Charlotte, NC, April 1993.

   

• W. Li, Voice and Data Integrated Service Token Ring Network, M.S. Thesis, December 1991.

   

• R. Sankar and S. Varada, Dynamic Protocol for Token Ring Network with Unbalanced Load, Proc. IEEE Southeastcon '92, Birmingham, AL, April 1992, pp. 10-13.

 

---

 

FDDI

Simulation modeling and performance analysis of networks including traditional local networks such as token-ring and high-speed networks such as FDDI have been conducted.

A simulation model to analyze the performance of FDDI MAC protocol has been developed and formulated a procedure to select and tune FDDI network parameters to obtain the best possible performance under different operating conditions. An analytical model to estimate the throughput of FDDI network with synchronous and asynchronous traffic was established and the boundary condition and stability of this network was studied.

Fault-tolerant design for dual-ring networks

A dual-counter rotating rings such as FDDI with reconfiguration capability can enhance the network fault-tolerance. In this project, an automatic failure isolation and reconfiguration methodology (FIRM) was developed for dual counter-rotating ring networks to detect, locate, and bypass physical failures. The strategy is to significantly improve the network reliability with maximum possible connectivity and without increasing the network length. A software implementation of the FIRM algorithm have been carried out.

 

---

For Details See the Relevant Publications

---

 

• Y.Y. Yang and R. Sankar, An Automatic Failure Isolation and Reconfiguration Methodology for Dual-Ring Networks, IEEE Network , Vol. 7, No.5, pp. 44-53, September 1993.

   

• Y. Y. Yang and R. Sankar, Analysis of Maximum Throughput for FDDI Network with Combined Synchronous and Asynchronous Traffic, Proc. Second International Conf. on Computer Communications and Networks (IC3N), San Diego, CA, June 1993, pp. 33-37.

   

• R. Sankar and Y. Y. Yang, An Automatic Failure Isolation and Reconfiguration Methodology for FDDI, Proc. International Conf. on Communications (ICC '92), Chicago, IL, June 1992, pp. 186-190.

   

• Y. Y. Yang and R. Sankar, Maximizing FDDI network performance by parameter tuning, Proc. Infocom '91, Miami, FL, Apr. 1991, pp. 1431-1439.

   

• Y. Y. Yang, Performance Modeling and Automatic Reconfiguration Methodology for Fiber Distributed Data Interface (FDDI) Network, Ph.D. Dissertation, June 1993.

   

• Y. Y. Yang, Modelling and Performance Analysis of Fiber Distributed Data Interface (FDDI), M.S. Thesis, July 1989.

   

• R. Sankar and Y. Y. Yang, Fault Tolerant and High Performance FDDI Network for Space Station Application, The Florida High Technology and Industry Council, February 1991, (Final Report, 61 pages).

 

---

 

Integrated Services Support in MAN-ATM Networks

The main objective is to support integrated services in a MAN (DQDB subnet) internetworked with ATM while satisfying the QoS requirements. An architecture is presented for the support of integrated services between two remote MAN users via ATM network. A dynamic bandwidth sharing scheme via renegotiation for applications with flexible bandwidth requirements is proposed.

 

---

For Details See the Relevant Publications

---

 

• K. Kidambi, R. Sankar, and J. Ottensmeyer, A QoS Mapping Scheme and a Model for MAN-ATM Inter(net)working, Computer Communications, Vol. 19, No. 3, pp. 276-286, March 1996.

   

• K. Kidambi, R. Sankar, and H. Kaur, A Dynamic Bandwidth Allocation Rule for Connection-oriented DQDB-ATM and Integrated Services Support, Proc. IEEE Globecom '95, Singapore, November 1995, pp. 383-387.

   

• K. Kidambi, R. Sankar, and J. Ottensmeyer, Internetworking MANs to ATM for Broadband Services Support, Proc. 20th IEEE Annual Conference on Local Computer Networks, Minneapolis, MN, October 1995, pp. 102-111.

   

• K. Kidambi and R. Sankar, MAN-ATM Internetworking and Integrated Services Support, Proc. IEEE Southeastcon '95, Raleigh, NC, March 1995, pp. 1-5.

   

• K. Kidambi, Bandwidth Allocation Using QoS Mapping and a Renegotiation Scheme for Integrated Services MAN-ATM Networks, Ph.D. Dissertation, May 1996.

 

---

 

Real-Time Network Performance Management

The objective is to develop and demonstrate the capability of applying modeling and simulation techniques to maintain the network integrity of a packet-switched network. A new technique called performance trending is introduced, in which key network parameters are monitored and network failure is predicted. Both hard failures in terms of physical breakdown and soft failures in terms of degradation in network performance level or quality of service (QoS) are considered for early detection so as to predict and prevent catastrophic failure of the network.

 

---

For Details See the Relevant Publications

---

 

• R. Sankar and R. Singh, Performance Trending for Management of Packet-Switched Networks, International Journal of Network Management, Vol. 4, No. 2, pp. 69-78, June 1994.

   

• R. Sankar and R Singh, Performance Trending for Management of High-Speed Packet-Switched Networks, Proc. IEEE Globecom '93 , Houston, TX, December 1993, pp. 1777-1781.

   

• R. Singh, Performance Management of Packet-Switched Networks, M.S. Thesis, July 1992.

 

---

 

 

---

Project Description - Digital Signal Processing

---

   

Speech Signal Processing

 

Robust Speech Recognition by Noise Immunization

Developed a robust speech recognition system using neural network. Using the multi-layer perceptron (MLP) neural network as a robust classifier and a modified backpropagation training algorithm, noise immunity is achieved for SNR levels down to 5 dB while maintaining high recognition accuracy. The use of noise immunization technique and the correlation of performance with the order of data presentation for network training are studied.

A word spotting system is developed to recognize the keyword 'collect' corrupted by white Gaussian noise in continuous speech.

 

Robust Recognition for Mobile Communication Applications

Developed robust speech recognition techniques for voice activated dialing for cellular phone in a car. Noise reduction techniques including linear and nonlinear spectral subtraction are implemented before modeling the speech parameters in the homomorphic domain. The FFT derived cepstral coefficients are liftered and then Mel-scale warped to generate the feature vector. Radial Basis Function (RBF) neural Network is used as the final classifier and its real-time performance is evaluated. Performance evaluation is carried out for both speaker-dependent and speaker-independent recognition across -5 dB to 25 dB SNR range with different front-end processing. The system is evaluated using NOISEX-92 and TIDIGITS noise and speech databases.

 

Speech Processing and Recognition Algorithms

Developed algorithms for pitch detection, endpoint detection, feature extraction, and several other processes involved in speech processing and recognition.

 

Speech and Speaker Recognition for Human ID and Biometric Applications

 

---

For Details See the Relevant Publications

---

 

• R. Sankar and S. Patravali, Noise Immunization Using Neural Net for Speech Recognition, Proc. IEEE International Conf. on Acoustics, Speech and Signal Processing (ICASSP), Adelaide, Australia, April 1994, Vol. II, pp. 685-688.

   

• S. Patravali, Robust Speech Recognition Using Neural Networks, M.S. Thesis, August 1995.

   

• R. Sankar and S. Patravali, Robust Speech Recognition by Noise Immunization Using Neural Network, Proc. Artificial Neural Networks in Engineering (ANNIE '93), St. Louis, MO, November 1993.

   

• H. Ruan and R. Sankar, Applying Neural Network to Robust Keyword Spotting in Speech Recognition Application, International Conference on Neural Networks (ICNN), Perth, Australia, November 1995, pp. 2882-2886.

   

• H. Ruan, Applying Neural Network to Robust Keyword Spotting in Speech Recognition Application, M.S. Project, April 1995.

   

• R. Sankar and N. Sethi, Robust Speech Recognition Techniques Using a Radial Basis Function Neural Network for Mobile Applications, IEEE Southeastcon '97, Blacksburg, VA, April 1997, pp. 87-91.

   

• N. Sethi, Evaluation of Noise Reduction Techniques and RBF Neural Network for Robust Speech Recognition in Mobile Applications, M.S. Thesis, December 1996.

   

• S. Varada and R. Sankar, Hardware Strategies for End Point Detection, Proc. IEEE Southcon, Ft. Lauderdale, FL, March 1995, pp. 163-167, (also accepted for The Fifth International Conf. on Signal Processing Applications & Technology - ICSPAT, Dallas, TX, October 1994).

   

• V. K. Sundaresan, S. Nichani, N. Ranganathan, and R. Sankar, A VLSI Hardware Accelerator for Dynamic Time Warping, Proc. 11th IAPR International Conf. on Pattern Recognition, The Hague, The Netherlands, August/September 1992, Vol. IV, pp. 27-30.

   

• V. K. Sundaresan, Software and Hardware Solutions for Dynamic Time Warping Algorithm, M.S. Thesis, December 1991.

   

• R. Sankar, Implementation of an Experimental Speaker-Independent Discrete Utterance Recognition System, Proc. IEEE International Conf. on Signal Processing, Beijing, China, October 1990, pp.445-448.

   

• M. E. Thompson, PC-Based Isolated Word Recognition System, M.S. Thesis, July 1990.

   

• R. F. Lorenzoni, Software Implementation of an LPC Based Isolated Word Recognition System, M.S. Thesis, July 1989.

   

• R. Sankar, A Pitch Extraction Algorithm for Voice Recognition Applications, Proc. 20th Southeastern Symposium on System Theory, Charlotte, NC, March 1988, pp. 384-387.

   

• R. Sankar, Experimental Evaluation of Structural Features for a Speaker-Independent Voice Recognition System, Proc. 20th Southeastern Symposium on System Theory, Charlotte, NC, March 1988, pp. 378-382.

 

---

 

Biomedical Signal Processing

 

Digital Mammography

A new mixed feature multistage false positive (FP) reduction method has been developed for improving the FP reduction performance. Eleven features were extracted from both spatial and morphology domains in order to describe the micro-calcification clusters (MCCs) from different perspectives. These features are grouped into three categories: gray-level description, shape description and clusters description. Two feature sets that focus on describing MCCs on every single calcification and on clustered calcifications, respectively, were combined with a back-propagation (BP) neural network with Kalman filter to obtain the best performance of FP reduction. First, 9 of the 11 gray-level description and shape description features were employed with back propagation neural network to eliminate all the obvious FP calcifications in the image. Second, the remaining MCCs will be classified into several clusters by a widely used criterion in clinical practice, and then the two cluster description features will be added to the first feature set to eliminate the FP clusters from the remaining MCCs. The new module combined existing tree-structure non-linear filter and back-propagation neural network with Kalman filter, the performance result of this new method was obtained using an image database of 100 real cases of patient’s mammographic images in H. Lee Moffitt Cancer Center imaging program.

 

Digital Image Standarization

As the soft copy reading and CAD in mammography become more and more important, the standardization of digital images becomes paramount. Telemammography and telemedicine requires the standardization for image characteristics, such as image resolution, bit-depth and intensity response. Soft copy reading and CAD in mammography are both dependent on the characteristics of the source of the digital data, either direct digital mammography or digitized screen-film mammography. An algorithm developed on images from one database may not perform as well on images from another database (with a different digitization). In this paper, we describe two methods based on a genetic algorithm and a nonlinear algorithm for standardization of digitized and digital mammography. The proposed standardization techniques are based on geometric and intensity transformations that are discovered using a set of calibration images. A set of transformation algorithm is used to search for the best standardization.

 

Lung Nodule Detection System

In this research, we develop a knowledge-based system for segmenting and labeling lung nodule on CT images. The system was developed in a blackboard environment that incorporates lung knowledge model, image processing model and inference engine. Lung model, which contains anatomical knowledge about lung in the form of semantic networks, is used to guide the interpretation process. The system works in a hierarchical structure, from large structures to the final nodule candidates, by focusing on the interested region step by step. The symbolic variables introduced to accomplish high-level inference, are defined by fuzzy confidence functions in lung model. Composite fuzzy functions are used to map between image and lung model objects. Anatomical lung segments knowledge is embedded in the system to direct 3-D validation of suspicious objects.

 

Prostate Boundary Detection

The objective of this research is to improve prostate boundary detection system by modifying a set of preprocessing algorithms including tree-structured nonlinear filter (TSF), directional wavelet transforms (DWT) and tree-structured wavelet transform (TSWT). A new advanced automatic edge delineation model for the detection and diagnosis of prostate cancer on transrectal ultrasound (TRUS) images is presented. The model consists of a preprocessing module and a segmentation module. The preprocessing module is implemented for noise suppression, image smoothing and boundary enhancement. The active contours model is used in the segmentation module for prostate boundary detection in two-dimensional (2D) TRUS images. Experimental results show that the preprocessing module improves the accuracy and sensitivity of the segmentation module greatly when a comparison made on the segmented images between those with preprocessing and those without preprocessing. It is believed that the proposed automatic boundary detection module for the TRUS images is a promising approach, which provides an efficient and robust detection and diagnosis strategy and acts as “second opinion” for the physician’s interpretation of prostate cancer.

 

Pulse Oximetry

Designed and implemented a more accurate and cost-effective portable pulse oximeter using spectral analysis. The pulse oximeter, a standard equipment in operating rooms, critical care units, and emergency health care, measures the percent oxygen saturation (SpO2) in hemoglobin.

The main objective and the resulting contributions of this research addressed two aspects of designing a portable pulse oximeter. The first was to show the use of spectral analysis to calculate SpO2 values is a practical solution. The second was to identify alternate methods to Weighted Moving Average algorithm that is currently used to compute these values. The digital signal processing algorithms using Fast Fourier Transform (FFT) and Discrete Cosine Transform (DCT) were designed and evaluted to provide higher accuracy and improved response.

 

EEG Signal Processing

Developed algorithms for automatic detection of transients in EEG and enhanced using classification techniques. The transients considered include spikes and spike and wave bursts which are abnormal phenomena associated with epileptic activity. The classification was enhanced using both patient-dependent and patient-independent analyses.

 

---

For Details See the Relevant Publications

---

 

• T. Rusch, R. Sankar, and J. Scharf, Signal Processing Methods for Pulse Oximetry, Computers in Biology and Medicine, Vol. 26, No. 2, pp. 143-159, March 1996.

   

• Biomedical variables section: Blood Chemistry (with T. Rusch), The Measurement, Instrumentation and Sensors Handbook, J. G. Webster (Editor), CRC Press, 1996.

   

• T. Rusch, J. Scharf, and R. Sankar, Alternate Pulse Oximetry Algorithms for SpO2 Computation, Proc. 16th Annual International Conf. of the IEEE Engineering in Medicine and Biology Society, Baltimore, MD, November 1994, pp. 848-849.

   

• T. Rusch, Implementation and Design of a Portable Pulse Oximeter Using Spectral Analysis, M.S. Thesis, November 1994.

   

• T. L. Rusch, R. Sankar, and J. E. Scharf, The Development of a Portable Pulse Oximeter for the Detection of Critical Hypoxemic Events in Non-surgical Patients, Group Technologies, Technical Report, December 1994.

   

• R. Sankar and J. Natour, Automatic Computer Analysis of Transients in EEG, Computers in Biology and Medicine, Vol. 22, No. 6, pp. 407-422, November 1992.

   

• R. Sankar and J. Goldstein, Computer-Aided Diagnosis of Epileptiform Transients in EEG, Proc. 21st IEEE Southeastern Symposium on System Theory, Tallahassee, FL, March 1989.

   

• J. D. Natour, Automatic Detection and Classification of Epileptiform Transients in EEG, April 1988.

   

• R. Sankar and J. Goldstein, Signal Processing and Pattern Recognition Approach to Transient Detection and Classification of EEG for the Diagnosis of Epilepsy, Proc. IEEE Southeastcon '87, Tampa, FL, April 1987, pp. 405-408.

   

---

Telemedicine

---

 

One of the goals of NCI to reach 80% of eligible women in mammography screening by 2000 yet remains a challenge. Recent medical reports reveal that while other types of cancer are experiencing negative growth, breast cancer is the only one with a positive growth rate. This is due to 1) examination process being a complex and lengthy one and 2) mammography not being available to the majority of women who live in remote sites. Currently, for mammography screening, women have to go to doctors or cancer centers annually while high-risk patients visit more often. To resolve these problems advanced networking technologies and signal processing algorithms can be used. Software modules can help detect true negatives (TN), while marking true positives (TP) for further investigation. Unavoidably, false negatives (FN) will be generated that are potentially life threatening. Inclusion of detection software improves the TP detection and hence reduces FNs drastically. Since TNs are the majority of examinations on a randomly selected population, this reduces the load on radiologists. High-speed communications can accelerate the required clinic-lab connection and make these algorithms readily available to radiologists. This expands diagnostics and treatment to the remote sites.

This research proposed ATMTN, an architecture for real-time, on-line screening, detection and diagnosis of breast cancer. ATMTN is a unique high-speed network integrated with automatic robust CAD/DSP methods for mass detection, Region of Interest (ROI) compression algorithms using DICOM 3.0 medical image standard. While ATMTN has the advantage of higher penetration into the women for cancer screening, it provides the diagnosis with higher efficiency, better accuracy, and potentially lower cost. The research goals involve: (1) Networking stations for telemammography to demonstrate, evaluate, and validate technologies and methods for delivering mammography screening services via high-speed (155 Mbytes/s) links, performing real-time network-transmitted, high-resolution mammograms for immediate diagnosis as a “second opinion” strategy. (2) Development of object-oriented compression methods for storage, retrieval, and transmission of mammograms. (3) Inclusion of detection algorithms for identification of normal images in different resolutions. (4) Resolving the compatibility issues between images from different equipment (DICOM standards) and (5) Optimization of the integrated ATMTN.

 

---

For Details See the Relevant Publications

---

 

 

---

Other Applications

---

 

Image Coding and Compression

The primary goal of this research is to develop a wavelet-based region of interest (ROI) coding for high-resolution medical imaging (16 bit X-ray mammogram). Number of algorithms was developed including

• An integer-to-integer shape adaptive discrete wavelet transform that is capable of coding the region losslessly.
• Lossless and lossy image compression methods for 16 bit x-ray images that can support region of interest (ROI) coding using Histogram-based Partial SPIHT and Partial SPIHT algorithms.
• A prioritized ROI coding that codes each region at a bit rate based on its priority, subjective and objective measures.

 

Image Quality Assessment Algorithms

A new novel algorithm for image quality assessment is proposed. First, a simple model of human visual system, consisting of a nonlinear function and a 2-D filter, processes the input images. This filter has one user-defined parameter, whose value depends on the reference image. In the next step the average value of locally computed correlation coefficients between the two processed images is found. This criterion is closely related to the way in which human observer assesses image quality. In the last step image quality measure is computed as the average value of locally computed correlation coefficients, adjusted by average correlation coefficient between the reference image and error image. This way the proposed measure differentiates between the random and signal-dependant distortion, which have different effects on human observer. Performance of the proposed quality measure is illustrated by examples involving images with different types of degradation.

This research involves frequency-domain-based image quality assessment algorithm for JPEG2000 compressed images. The algorithm segments the frequency response of the original and compressed images into four bands and then evaluates the normalized means of contrast sensitivity function (CSF) weighted frequency magnitudes and normalized means of CSF weighted absolute error between original and compressed images within each band to relate to the visual quality of the compressed images. The image quality is assessed by evaluating three visual characteristics for tested images, texture, edges and overall quality, and classified according to a numerical quality scale.

Unlike many objective quality assessment metrics, the proposed algorithm, which incorporates the contrast sensitivity function modeling the behavior of the human visual system at various spatial frequencies, relates to the perceptual measures by giving information about the content of the tested images. The experimental results show satisfactory estimate of the actual image quality.

 

Adaptive Signal Processing

Existing opportunities in advanced interceptor technology and satellite guidance programs, requiring higher signal processing speeds and lower noise environments, are demanding Ring Laser Gyro (RLG) based Inertial Systems reduce initialization and operational data latency as well as correlated noise magnitudes. Existing signal processing algorithms are less than optimal when considering these requirements. Research of real-time adaptive signal processing algorithms for use in RLG based inertial systems will help to understand the trade-off in correlated noise magnitude, organizational complexity, computational efficiency, rate of convergence, and numerical stability. Adaptive filter structures selected will directly affect meeting inertial system performance requirements for data latency, residual noise budgets and real time processing throughput.

The primary goal of this research project is to develop adaptive noise cancellation algorithms for RLG based inertial systems in a variety of military and commercial space applications. Of particular significance is an attempt to identify an algorithm that will reduce the correlated noise components to the theoretical limit of the RLG sensor itself. This would support a variety of applications in the low noise space environments that the RLG based inertial systems are beginning to find promise for such as advanced military interceptor technology and commercial space satellite navigation, guidance and control systems.

Continuing research will explore the development and implementation of real-time adaptive filters including Least Mean Square (LMS) and Recursive Least Square (RLS) structures for the purpose of evaluating performance tradeoffs such as convergence rate, effectiveness of correlated noise cancellation, organizational complexity and computational efficiency. Development of such real-time adaptive filters for use in dithered ring laser gyro based inertial systems will insure an optimal selection of candidate structures for a wide variety of commercial and military applications.

 

---

For Details See the Relevant Publications

---