摘要
为了在噪声环境中实现准确的信噪比估计,提出了一种高效准确的盲信噪比估计方法,该方法将图信号分析方法用于语音盲信噪比估计。研究发现,若样本量有限,与噪声相比语音信号转换所得的图具有更低稀疏性。基于这一发现,利用语音活动检测器获取语音活动区域,将区域内信号转换为图并计算其稀疏度,利用稀疏度与信噪比关系估计语音信号的信噪比。实验结果表明,与现有的方法相比,该算法能更准确地估计信噪比,尤其在信噪比为-14 dB到-4 dB环境下表现出优越性。
To achieve accurate signal-to-noise ratio(SNR)estimation in noisy environments,an efficient and accurate blind SNR estimation method is proposed.This method uses the graph signal analysis method for speech blind SNR estimation.It is found that if the sample size is limited,the graph obtained by speech signal conversion has lower sparsity than that obtained by noise.Based on this finding,the voice activity detector is used to identify the voice activity area,transforms the signal within the region into a graph and calculates its sparsity,and estimates the SNR of the speech signal by analyzing the relationship between sparsity and SNR.The experimental results show that the proposed algorithm can estimate the SNR more accurately compared with the existing methods,especially in the SNR range of-14 dB to-4 dB.
作者
陈海贞
闫坤
宁振杰
王对强
CHEN Hai-zhen;YAN Kun;NING Zhen-jie;WANG Dui-qiang(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004,China;Satellite Navigation Positioning and Location Service National&Local Joint Engineering Research Center,Guilin University of Electronic Technology,Guilin 541004,China;School of Artificial Intelligence,Hezhou University,Hezhou 542899,China)
出处
《计算机工程与设计》
北大核心
2025年第8期2170-2177,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(62101147)
广西自然科学基金项目(桂科2020GXNSFAA159146)
广西创新驱动发展专项基金项目(桂科AA21077008)
教育部重点实验室基金项目(CRKL190108)。
关键词
盲信噪比估计
非平稳性
图信号分析方法
梅尔能量
语音活动检测器
图构造
图稀疏性
blind signal-to-noise ratio estimation
non-stationarity
graphical signal analysis method
mel energy
voice activity detector
graph construction
graph sparsity
