采煤工作面水流声的降噪算法研究
Research on denoise algorithm of water flow sound in coal mining face
李传兵 1李思毅 2程瑶2
作者信息
- 1. 重庆理工大学 机械工程学院,重庆 400054;朗德科技(重庆)有限公司,重庆 400054
- 2. 重庆理工大学 机械工程学院,重庆 400054
- 折叠
摘要
采煤工作面水害事故发生前伴有水流声出现,为减小常见机械设备噪声对水流声识别的影响,提出了一种改进小波阈值降噪法.为了补偿水流声在采集过程中高频信号的损失,对其进行了预加重、分帧、加窗等预处理;采用小波软、硬阈值函数折衷法对阈值函数进行改进,避免了信号处理后带来的附加振荡问题;在固定阈值中引入了分解层数j,变化的阈值可最大程度分离水流声信号与噪声信号.结果表明,改进小波阈值降噪法相较于谱减法、子空间法和小波阈值法的硬阈值函数,输出信噪比提高了1 dB以上,降噪性能提升了15％左右;相较于小波阈值法的软阈值函数,当输入信噪比大于0 dB时,输出信噪比提高1 dB以上,可有效降低水流声信号中的机械设备噪声干扰.
Abstract
The water damage accident at the coal mining face is preceded by the sound of water flowing. To reduce the influence of noise interference of common mechanical equipment on the sound recognition of water flow, this paper proposes an improved wavelet threshold noise reduction method. To compensate the loss of high-frequency signals in the process of water flow sound acquisition, pre-emphasis, framing, and windowing are pre-emphasized. The compromise method of wavelet soft and hard threshold functions is employed to improve the threshold function, avoiding the additional oscillation problem caused by signal processing. The number of decomposition layers j is introduced into the fixed threshold, and the variable threshold maximizes the separation of the water flow acoustic signal from the noise signal. Our results show the hard threshold function improves the output signal-to-noise ratio by more than 1 dB and reduces the noise by about 15％ when compared with those of spectral subtraction, subspace method and wavelet threshold method. Compared with the soft threshold function of the wavelet threshold method, the output signal-to-noise ratio increases by over 1 dB when the input signal-to-noise ratio ≥0 dB. The improved wavelet threshold noise reduction method effectively reduces the noise interference of mechanical equipment in the water flow acoustic signal.
关键词
水流声/降噪/小波变换/阈值函数/信噪比Key words
water flow sound/noise reduction/wavelet transform/threshold function/signal-to-noise ratio引用本文复制引用
基金项目
重庆市科委基础与前沿研究一般项目(cstc2016jcykA0497)
出版年
2024
国家科技期刊平台
NSTL
万方数据