摘要
针对广义旁瓣相消器(GSC)中固定波束形成器在复杂环境下抑制旁瓣干扰和处理非平稳语音信号时存在局限性问题,提出了一种基于参数化波束形成器的改进GSC语音增强方法。该方法通过动态调节机制,在延迟求和波束形成器与超指向波束形成器之间进行灵活权衡与调节,有效抑制旁瓣干扰,增强了GSC在复杂声学环境中的鲁棒性与适应性。此外,引入互相关系数来调节自适应滤波器权重更新步长,有效应对语音信号变化导致的过减问题,提升了在非平稳语音信号中的处理精度。在MATLAB环境下开展仿真实验,针对Babble噪声、音乐噪声和白噪声环境,对比传统GSC和采用均方算法的GSC,从三维波束方向图、不同背景噪声及参数条件下的降噪效果、互相关系数作用效果等方面进行评估,并利用分段信噪比(SNR)和语音质量感知评估(PESQ)等指标量化分析。结果显示,改进方法在降噪性能和语音清晰度上优势显著。在Babble噪声、音乐噪声、白噪声环境中,分段信噪比分别提升至11.02、6.14和10.33 dB,PESQ值分别提升至3.65、3.20、3.25,并可通过调节参数实现不同噪声环境下的最佳降噪效果,有力验证了该方法在复杂声学环境中的有效性与优越性。
To address the limitations of the fixed beamformer in the generalized sidelobe canceller(GSC)in suppressing sidelobe interference and processing non-stationary speech signals in complex environments,this paper proposes an improved GSC-based speech enhancement method utilizing a parameterized beamformer.The proposed method employs a dynamic tuning mechanism to flexibly balance and adjust between the delay-and-sum(DS)beamformer and the super-directive(SD)beamformer,effectively suppressing sidelobe interference and enhancing the robustness and adaptability of the GSC in complex acoustic environments.Furthermore,a cross-correlation coefficient is introduced to regulate the step size of the adaptive filter weight update,mitigating the over-attenuation issue caused by variations in speech signals and improving the processing accuracy for non-stationary speech signals.Simulation experiments were conducted in MATLAB to evaluate the performance of the proposed method under various noise conditions,including Babble noise,music noise,and white noise.The performance of the proposed method was compared with that of the traditional GSC and the GSC with the least mean square(LMS)algorithm.The evaluation was carried out from multiple perspectives,including 3D beam patterns,noise reduction effects under different background noise and parameter conditions,and the effectiveness of the cross-correlation coefficient.Quantitative analysis was performed using performance metrics such as segmental signal-to-noise ratio(SNR)and perceptual evaluation of speech quality(PESQ).The results demonstrate that the proposed method significantly outperforms the traditional GSC in terms of noise reduction performance and speech clarity.In environments with Babble noise,music noise,and white noise,the segmental SNR improved to 11.02,6.14,and 10.33 dB,respectively,while the PESQ values increased to 3.65,3.20,and 3.25,respectively.By adjusting the parameters,the proposed method achieves optimal noise reduction effects in different noise environments,validating its effectiveness and superiority in complex acoustic scenarios.
作者
张传营
赵景玉
刘扬
卜凡亮
Zhang Chuanying;Zhao Jingyu;Liu Yang;Bu Fanliang(School of lnformation and Network Security,People’s Public Security University of China,Beijing 102600,China)
出处
《电子测量与仪器学报》
北大核心
2025年第5期125-133,共9页
Journal of Electronic Measurement and Instrumentation
基金
中国人民公安大学双一流创新研究项目(2023SYL08)资助。
关键词
麦克风阵列
波束形成器
广义旁瓣相消器
参数化
互相关系数
microphone array
beamformer
generalized sidelobe canceller
parameterization
cross-correlation coefficients
作者简介
张传营,2022年于河北科技大学获得学士学位,现为中国人民公安大学硕士研究生,主要研究方向为阵列及语音信号处理。E-mail:962409974@qq.com;赵景玉,2021年于东北大学获得学士学位,现为中国人民公安大学硕士研究生,主要研究方向为阵列信号处理与自动化控制。E-mail:1754633076@qq.com;刘扬,2022年于山东科技大学获得学士学位,现为中国人民公安大学硕士研究生,主要研究方向为阵列信号处理与安全防范技术。E-mail:1419311746@qq.com;通信作者:卜凡亮,1993年于成都电子科技大学获得学士学位,1996年于西安交通大学获得硕士学位,2000年于西安交通大学获得博士学位,现为中国人民公安大学安全防范系教授,主要研究方向为信息与信号处理。E-mail:bufanliang@sina.com。
