基于三维协同优化的低功耗FIR滤波器设计方法研究

Finite impulse response (FIR) filter is a fundamental building block in many digital signal processing (DSP) systems. They are widely used in a lot of DSP applications. With the development of portable electronic devices, low-power FIR filter design is becoming increasingly important. Most of the existing low-power FIR filter design techniques focus on optimizing filter coefficients and common subexpression sharing. While ignored by most of the existing literatures, the optimization of filter structure is crucial for power minimization. In this project, a 3D optimization approach will be explored to minimize the power consumption of FIR filters. In this 3D optimization approach, the filter coefficients, the direct-transpose hybrid form filter structure as well as the common subexpression sharing will be jointly optimized in a single algorithm. The power consumption of hybrid form FIR filters will be analyzed and a power model will be proposed. Based on that, a filter design algorithm with embedded common subexpression sharing and hybrid structure optimization will be developed. In the proposed algorithm, the hybridization of direct and transposed structures as well as the common subexpression sharing will be dynamically updated during the search of filter coefficients. Through this project, the low-power FIR filter design theory will be enriched. This project will also boost the development of research areas such Internet-of-Things, wearable devices as well as mobile communication.

作为最重要的组成单元之一，数字FIR滤波器被广泛应用于各类数字信号处理系统中。随着电子设备朝着小型化、便携化的方向发展，低功耗FIR滤波器设计的重要性正日益显现。传统的低功耗FIR滤波器设计主要基于转置型结构进行滤波器系数和共同子式共享优化，却往往忽略了滤波器结构对功耗的影响。本项目拟从FIR滤波器整体设计出发，以功耗为目标，对滤波器的系数、直接-转置结构混合以及共同子式共享进行三维协同优化。本项目将首先对直接-转置混合型FIR滤波器进行功耗分析并建模。在此基础上，设计嵌入直接-转置混合优化和共同子式共享优化的滤波器系数设计算法，在搜索滤波器系数的同时动态地优化滤波器结构中直接与转置的混合以及共同子式的共享，从而优化整个滤波器电路的功耗。通过本项目研究，将能够进一步完善低功耗FIR滤波器设计理论和方法，同时对诸如物联网、可穿戴设备、移动通信等重要领域产生重要的促进作用

作为最重要的组成单元之一，数字FIR滤波器被广泛应用于各类数字信号处理系统中。随着电子设备朝着小型化、便携化的方向发展，低功耗FIR滤波器设计的重要性正日益显现。本项目针对低复杂度滤波器，从滤波器系数稀疏设计、级联滤波器结构和系数稀疏设计、滤波器系数可重构等角度，对低复杂度滤波器和滤波器组进行了相应的研究。基于上述研究，提出了多种不同的滤波器系数和结构优化设计方法，从而优化整体滤波器电路的设计。同时，本项目针对一些典型的基准滤波器进行了流片测试。实验结果表明，本项目所提出的算法和结构，可以有效地降低滤波器的复杂度，从而减少整体电路的功耗。通过本项目研究，将能够进一步完善低功耗FIR滤波器设计理论和方法，同时对诸如物联网、可穿戴设备、移动通信等重要领域产生重要的促进作用。