摘要
针对多噪声环境下非平稳声信号的去噪问题,为提高信号识别精度,利用小波变换具有多分辨分析的特点,提出一种改进的小波阈值去噪方法,并对阈值的选取和阈值函数的选择做出改进。利用对数函数的特性拟合分解层变量对噪声的影响,提出一种动态阈值计算方法,保证了阈值在随分解层数增加时会随之变化,优化了采用固定阈值带来的信号过度滤波问题;利用指数函数的特性设计了一个阈值函数,解决了采用传统阈值函数带来的信号重构偏差大的问题;最后,以信噪比和均方根误差为性能指标进行实验,对比了不同阈值和阈值函数下的去噪精度。结果表明,文中设置的动态阈值计算方法在去噪时误差最小,更加逼近目标信号,所设计的方法提高了信号识别精度,提升了小波变换的实用价值。
In allusion to the denoising of non⁃stationary acoustic signal in a multi⁃noise environment,an improved wavelet threshold method is proposed and the selection of threshold and threshold function are improved to raise the accuracy of signal recognition by utilizing the characteristics of multi⁃resolution analysis of wavelet transform.In one aspect,the characteristics of the logarithmic function is used to fit the influence of decomposition layer variables on noise.And then,a dynamic threshold calculation method is proposed to ensure that the threshold will change as the number of decomposition layers increases,and the excessive signal filtering caused by the use of a fixed threshold is optimized.In another aspect,a threshold function is designed based on the characteristics of the exponential function,which gets rid of the large signal reconstruction deviation caused by the traditional threshold function.Experiments were carried out.In the experiment,signal⁃to⁃noise ratio and root⁃mean⁃square error were taken as performance indicators,and the denoising accuracy at different thresholds and threshold functions were contrasted.The result shows that the set dynamic threshold calculation method has the smallest error when denoising,and it is closer to the target signal.Therefore,the designed method can improve the accuracy of the signal recognition and enhance the practical value of wavelet transform.
作者
祖丽楠
刘志远
生宁
ZU Linan;LIU Zhiyuan;SHENG Ning(College of Automation and Electronic Engineering,Qingdao University of Science and Technology,Qingdao 266100,China)
出处
《现代电子技术》
2022年第11期35-40,共6页
Modern Electronics Technique
基金
国家自然科学基金青年科学基金项目(61703223)。
关键词
滤波
小波变换
非平稳声信号
小波阈值
阈值函数
信噪比
重构信号
wave filtering
wavelet transform
non⁃stationary acoustic signal
wavelet threshold
threshold function
signal⁃to⁃noise ratio
reconstructed signal
