Yongxin Zhu

Implementing Contourlet Transform for Medical Image Fusion on a Heterogenous Platform

The classic wavelet transform has been applied in registration and fusion of medical images for decades. An extension namely, contourlet transform recently indicated its advantages in image fusion with better efficiency in multi-resolution and multi-direction representation and calculation. However, the even higher computational complexity it requires turns out to be a disastrous concern in embedded applications. In this paper, we implement an acceleration system on a heterogeneous TI Da Vinci dual core processor consisting of an ARM processor core and a DSP processor core. The ARM core controls the fusion procedure by extracting the luminance bits and invoking the DSP core to carry out the time-consuming part of the contourlet transform. The partitions of tasks are determined after the program is analyzed and profiled at functional level to make full use of the computational capability of the heterogeneous platform. Initial measured improved performance results are obtained and analyzed with projected further improvements.

A Hybrid Anti-Collision Algorithm for RFID with Enhanced Throughput and Reduced Memory Consumption

In order to solve the transponder collision problem in a RFID system, this paper proposes a hybrid algorithm which combines strengths of existing high-performance algorithms while avoiding their major drawbacks. Experiments show this hybrid algorithm uses fewer time slots and less total communication time compared to adaptive slot-count algorithm based on ALOHA as well as enhanced anti-collision algorithm based on binary tree, both of which are superior and well accepted algorithms in literature. Meanwhile, high computational intensity of adaptive slot-count algorithm is significantly relieved, and memory consumption of enhanced anti-collision algorithm is reduced to a negligible level. Results from simulations under large number of transponders and long transponder IDs, which is the case in real RFID applications, suggest the advantage of this hybrid algorithm is prominent.

Statistical estimation and evaluation for communication mapping in Network-on-Chip

In order to exploit the advantages in on-chip communication introduced by Network-on-Chip, many optimization algorithms have been proposed for a joint optimization on power and performance in communication mapping and routing. However, the optimality of solutions relative to these algorithms has been neglected in previous studies. To this problem, this paper proposes an early estimating approach to evaluate the optimality of the solutions for the first time. This approach is based on a statistical property that the overall solutions in solution space conform to a quasi-Gaussian distribution, which can be previewed by two parameters with a computation complexity of O(n^4) as presented. The generality of our proposed approach makes itself extensible to other on-chip network options. Experiments on real and synthetic application benchmarks demonstrate an average error ratio less than 7% which tends to be even smaller when problem scales up. These results validate our early estimating approach on optimality evaluation as credible and efficient to boost its utility in the promising Network-on-Chip design.

Statistical Estimation for Total Communication Load in Application-Specific Network-on-Chip

Network-on-Chip design takes optimizing the specific communication requests for target topology as one of its synthesis stages. Instead of finding the optimal result, this paper first addresses the problem of evaluating the various solutions of different algorithms. To obtain such an evaluation, we introduce a statistical approach for approximating the solution space by the observation that all the solution values are Gaussian distributed. An efficient method for calculating the Gaussian parameters with a complexity of O(n4) at most is also presented, which is good at solving large scale problems. The superior fidelity of the approximation is validated by real application benchmarks as well as more synthetic ones, with the average error ratio within 7% and even smaller for larger scale problems. Our approach is efficient and creditable which will boost its utility in the promising Network-on-Chip designs.

Software Pipelining with Minimal Loop Overhead on Transport Triggered Architecture

On transport triggered architectures (TTAs) featuring huge scheduling freedom, parallelism is exploited at not only operation level, but also data transportation level. Software pipelining, an aggressive compiler optimization scheme for exploiting instruction level parallelism across loop iterations, has been studied extensively. However, only few efforts were focused on software pipelining on TTAs. In these existing works, intuitive yet less efficient methods were used, namely either modulo scheduling algorithm with some heuristics or parallel language to implement software pipelining on TTA. We propose a new software pipelining method on TTAs in order to fully evaluate the scope of scheduling freedom of TTA and take advantage of it. In this paper, we formulate the problem of constructing a resource constrained rate-optimal software pipelining with minimal loop overhead on TTAs as an integer linear programming (ILP) problem. The formulated problem is solved with GNU Linear Programming Kit (GLPK). We apply our approach to major loops in Livermore loop benchmarks. Comparing with the previous schedulers implemented with modulo scheduling algorithm, our ILP approach creates schedules which bring significant performance enhancement to applications on TTA.

The Design of a Wireless Power Transmission Mechanism for Locomotion in Active Medical Inspection MEMS

The minimally invasive inspection based on medical MEMS has been in clinical use since the Israel invention of a passive capsule endoscope, M2A, followed by similar capsule endoscopes made in Japan and Korea. Due to the need of external control, active medical MEMS containing autonomous locomotion started to appear with power cable connections to replace in sufficient batteries. Unfortunately, power cables cannot be used in medical practice. As such, wireless power transmission for active MEMS is a must for interests of both researchers and doctors. In this paper, we present a design of a wireless power transmission mechanism to drive a micro brushless DC motor whose small size fits MEMS. Though the power transmission is done at a working frequency of RFID, the power supply system in this paper to drive a motor witha much larger current is different from the one in RFID tags. The power transmission system consists of an antenna, an impedance matching network circuit, a rectifier circuit cum voltage multiplier, and a LDO (low drop out voltage regulator). Both mathematical analysis and simulations based EDA tools are carried out to verify the system design. The RF part of the system is simulated with Agilent ADS tools, while the DC part, i.e. the LDO based on MOSFETs, is simulated by Cadence Spectre with a TSMC 0.18um technology library. Both simulation results and mathematical analysis show that the micro motor is able to receive sufficient power to work in vivo in real time.

On Driving Micro Motors for MEMS with a Micro Driver Based on Three-Phase Sinusoidal Signals

Micro electromechanical system (MEMS) technologies, which have been widely used in robotics and automation, are being ported into medical embedded systems. A key change MEMS brings is the mechanic locomotion to physically move the embedded systems, which is usually actuated by a micro motor. However, classic motor driver and control cannot be used to drive the micro motor due to its tiny size and unique motor mechanisms. This paper presented a micro three-phase sinusoidal signal driver for micro brushless motor. The driver enables precise adjustment to the speed and output torque of the micro motor without any feedback. To explain the physical principles of the driver, we first analyze the compound Magnetic Motive Force (MMF) of three-phase sine-wave voltage. Then we describe the traditional way to drive the micro motor. After explaining the theoretical logic, we present the design details of the driver. After verifying the design with simulation, we implement the driver with physical circuits whose measured results are also shown. We believe that the correct design and implementation of the driver will enable active movements of embedded systems in medical environment.

Design and Implementation of a Cordless Power Supply System for Pervasive Medical Devices

Pervasive medical devices are able to provide diagnostic information inside intangible in vivo organs. For better inspection, additional power besides cell batteries is required by doctors to enable locomotion and active control of devices such as capsule endoscope. We present the design and implementation of a cordless power supply system based on 900MHz RF signals. It consists of an antenna, an impedance matching network circuit, a rectifier circuit cum voltage multiplier, and a LDO (low drop out voltage regulator). After matching analytical results and simulation ones, we further implement the power supply system in a PCB and measure it with Agilent spectrum analyzer and Tektronix oscilloscope. It is also interesting that wireless signals from any 900M GSM mobile phone talking nearby can be absorbed and converted into enough power by the system. It is believed to enable the external control and the locomotion of active endoscopes as well as other power hunger pervasive medical devices.

Adaptive robust dynamic balance and motion control of mobile wheeled inverted pendulums

Adaptive robust dynamic balance and motion control are considered for mobile wheeled inverted pendulum, in the presence of parametric and functional uncertainties. Based on Lyapunov synthesis, the proposed control mechanisms using physical properties of wheeled inverted pendulums ensure that the outputs of the system track the given bounded reference signals within a small neighborhood of zero, and guarantees semi-global uniform boundedness of all the closed loop signals. Simulation results are presented to verify the effectiveness of the proposed adaptive robust control.

Evaluation of Partitioning Methods for Stream Applications on a Heterogeneous Multi-core Processor Simulator

In this paper, we explore coordination of multiple cores in a single processor to achieve better performance and lower real time power consumption for single multimedia streaming application, as opposed to multiple independent applications executing in a parallel fashion. Two strategies of application partitioning, i.e. partitioning by data or function, are evaluated by implementing an audio application, which is to encode audio signals in G.721 protocol, on two partitions. To execute these partitioned tasks, we improve a multi-core processor simulator, whose performance, power metrics were verified before, by enabling real time power consumption. The coordination of multiple cores is assured with explicit inter-core communication which is also enabled in the simulator. It is interesting to find that partitioning by function outperforms partitioning by data in terms of both performance and power consumption. The observation, which should be valid for most of streaming applications, would be a good reference for system architects. The implementation in the paper also underscores the importance of and the need for coordinating multiple cores on a single processor via inter-core communications.