Low-resource FPGA Realization of Acoustic Feedback Cancellation for Hearing Aids

Read full text Summarise with VIRAT
Authors: Ali A. Mustafa, Ivan A. Hashim, Sabah Abdul-Hassan Gitaffa
Cited by: 0
References: 29

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Acoustic feedback is a significant problem in hearing aid systems, because it reduces the maximum attainable amplification and compromises the quality of the audio.In practical implementations of hearing aids, this challenge is further complicated by limited hardware resources and stringent real-time processing capability.These constraints provide the motivation for low-complexity and hardware-efficient acoustic feedback cancellation (AFC) solutions.Therefore, in this paper, a comprehensive study of the low-complexity AFC systems that can be used in hearing aid applications is presented.Starting from a baseline NLMS based AFC architecture, several optimization strategies are explored to minimize the computational complexity, hardware resource utilization and power consumption.The first approach implements the normalization operation of NLMS algorithm by using a CORDIC divider for implementation.Subsequently, a ROM-based normalization technique and a systematic fixed-point wordlength optimization technique are proposed to further improve implementation efficiency.In addition, an LMS-based AFC system is used to eliminate the use of normalization-related hardware altogether.Probe noise injection is included to reduce signal correlation and estimation bias.The proposed designs are evaluated by using MATLAB Simulink and implemented on Spartan-3A DSP FPGA board using the Xilinx ISE 14.7 design suite.The LMS-based AFC method gives the ERLE of 12.267 dB, PESQ score of 4.273, and added stable gain of 15.07 dB, which shows good feedback suppression and good speech quality.From a hardware perspective, the LMS-based design only requires 665 slices, 403 slice flip-flops, and 541 four-input LUTs, 22 DSP48 blocks, and has a maximum operating frequency of 10.415 MHz and a total power consumption of 0.960 W, which is the most hardware and power efficient solution among the evaluated approaches.Overall, the results confirm that the LMS-based AFC architectures offer an attractive trade-off between performance and implementation efficiency, making it suitable for real-time, resource-constrained hearing aid systems.

Fields of Study & Subjects

Advanced Adaptive Filtering Techniques Speech and Audio Processing Digital Filter Design and Implementation

Publication details

Journal / source
International journal of intelligent engineering and systems
Volume
19
Issue
4
Pages
847–860
Publication date
23 March 2026
Language
EN