580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can optimization techniques improve the design and stability of 2D nearly linear-phase IIR digital filters?
Designing nearly linear-phase two-dimensional (2D) IIR digital filters remains a significant challenge due to inherent nonlinearities and stability constraints. Optimization techniques, especially nonlinear constrained optimization, have been a focus to balance magnitude, phase, and stability requirements, enabling practical applications in image and biomedical signal processing where phase fidelity is critical.
Key finding: This work formulates 2D IIR filter design as a nonlinear constrained optimization problem, minimizing group delay deviation in the passband under magnitude and stability constraints. Using the interior-point method and a... Read more
Key finding: This study compares multiple evolutionary computation (EC) algorithms for the design of 2D-IIR filters, which are known for their unstable transfer functions and multimodal error surfaces. It demonstrates that EC methods,... Read more
Key finding: The paper introduces an interval size approach to analyze frequency response variations of IIR filters under finite precision coefficient quantization, particularly fixed-point arithmetic. It demonstrates how quantization... Read more
2. What are the effective methods for designing low-power and energy-efficient digital IIR filters using approximate computing?
Energy efficiency is critical in digital IIR filter implementations, especially for embedded and edge devices. Approximate computing, involving approximate multipliers and coefficient quantization, emerges as a viable strategy to reduce power consumption while maintaining acceptable filter performance, particularly in widely used applications like acoustic A-weighting filters.
On the Design of an Energy Efficient Digital IIR A-Weighting Filter Using Approximate Multiplication
Key finding: This paper presents a sixth-order digital IIR A-weighting filter design using approximate odd radix-4 multipliers to reduce hardware power consumption. The design achieves up to 70% energy reduction (power-delay product) with... Read more
Key finding: This work proposes a high-performance 2D digital filter architecture combining the Winograd filtering method with Residue Number System (RNS) arithmetic to reduce multiplication complexity and enhance parallelism. Although... Read more
Key finding: The paper develops an analytical framework for fixed-point IIR filter implementation, emphasizing customized bit widths for coefficients and internal signals to minimize overflow and quantization errors. This allows... Read more
3. How can adaptive and complex IIR filter structures enhance narrowband signal processing and filter bank designs?
Adaptive IIR filter architectures and complex coefficient filter banks have been investigated for improved signal analysis, harmonic tracking, and narrowband interference cancellation. They offer low sensitivity to coefficient quantization, fast convergence, and hardware-efficient implementations suitable for dynamic and noisy signal environments, including communications and speech enhancement.
Key finding: The paper introduces a novel adaptive complex digital filter structure with low magnitude sensitivity and independently tunable center frequency and bandwidth. Using LMS adaptation, it achieves rapid convergence and high... Read more
Key finding: This research develops cascaded-resonator (CR) based IIR filter banks, which are more hardware-efficient than FIR counterparts for spectral decomposition. The approach uses complex coefficients to realize analytic signals... Read more
Key finding: The authors propose novel systolic architectures for one-dimensional and two-dimensional delay least-mean-square (DLMS) adaptive IIR digital filters that optimize hardware regularity and throughput. By using tree-systolic... Read more