Unified Theory for Synthesis of Optimal Structures of Next-Generation Digital Filters Based on Stochastic Computation

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K06049
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Communication/Network engineering
• Research Institution: Tohoku University
• Principal Investigator: Koshita Shunsuke 東北大学, 工学研究科, 助教 (70431533)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 電気電子工学 / 通信・ネットワーク工学 / 制御工学 / 信号処理 / ディジタルフィルタ / 演算精度 / 確率的演算

Outline of Final Research Achievements

This research project has established theoretical approaches on unified synthesis of filter structures that achieve higher arithmetic precision of stochastic computation-based next-generation digital filters than conventional methods. First, a unified synthesis method has been developed for stochastic FIR filters, where the method makes use of the deterministic and stochastic hybrid computation. Next, two synthesis methods have been developed for stochastic IIR filters to improve arithmetic precision. The proposed methods respectively make use of the scaled/cascaded structure and the state-space equation. Furthermore, hardware implementation using the proposed theory has been conducted and it has been demonstrated that the proposed method can attain at most 40dB improvement in filter characteristics.

Free Research Field

工学