摘要
针对强噪声背景下综放开采过程中垮落煤矸难以识别问题,提出了一种融合低级听觉特征Mel频谱和高级听觉特征听觉神经递质发放率的煤矸识别方法。首先,根据煤矸垮落冲击液压支架尾梁声音信号频谱特点,基于听觉神经滤波器组模型构建了适用于煤矸识别任务的听觉模型;然后,利用听觉模型对煤矸垮落声音信号进行分析,获得听觉神经递质发放率;再次,将听觉神经递质发放率与通过Mel频谱提取的峰值特征进行融合,得到煤矸声音听觉感知图;最后,基于所构建的听觉感知图,利用ConvNeXt模型进行煤矸识别。试验结果表明,采用融合听觉特征的煤矸识别方法在不同信噪比下均具有较高的识别准确率;其优越性在背景噪声较大的工况下(信噪比为-5 dB)尤为明显,准确率仍能达到91.52%,显著优于以低级听觉特征和频谱图作为识别特征和利用时频域特征结合机器学习的煤矸识别方法,验证了融合听觉特征的煤矸识别方法对噪声具有优越的鲁棒性。
Aiming at the problem that it is difficult to recognize the caving coal gangue in the process of fully mechanized caving mining under the background of strong noise,a coal and gangue recognition method fusing low-level auditory feature Mel spectrum and high-level auditory feature auditory neurotransmitter firing rate wasproposed.Firstly,according to the frequency spectrum characteristics of the sound signal of the tail beam of collapsed coal and gangue impact hydraulic support,an auditory model suitable for the coal gangue recognition task wasestablished based on the auditory neural filter bank model.Then,the auditory model wasused to analyze the sound signal of collapsed coal and gangue to obtain auditory neurotransmitter firing rate.Afterwards,the auditory neurotransmitter firing rate wasfused with the peak feature extracted by Mel spectrum to obtain the auditory perception diagram of coal and gangue sound.Finally,coal and gangue were recognized with the ConvNeXt model based on the fusion auditory features constructed.Experimental results show that the proposed coal and gangue recognition method with fusion auditory features has high recognition accuracy under different signal-to-noise ratios,and its superiority isparticularly evident under the condition of large background noise(signal-to-noise ratio of-5 dB),with accuracy reaching 91.52%,which issignificantly superior to the method using low-level auditory features and spectrum as recognition features and using time-frequency domain features combined with machine learning,verifying the robustness of the proposed method to noise.
作者
杨政
王世博
饶柱石
杨善国
杨建华
刘送永
刘后广
YANG Zheng;WANG Shibo;RAO Zhushi;YANG Shanguo;YANG Jianhua;LIU Songyong;LIU Houguang(School of Mechatronic and Electrical Engineering,China University of Mining and Technology,Xuzhou 221116,China;Jiangsu Province and Education Ministry Co-sponsored Collaborative Innovation Center of Intelligent Mining Equipment,Xuzhou 221116,China;State Key Laboratory of Mechanical System and Vibration,Shanghai Jiao Tong University,Shanghai 200240,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2024年第8期136-144,共9页
Journal of Vibration and Shock
基金
国家自然科学基金资助项目(52274162,52275296)
江苏高校优势学科建设工程资助项目(PAPD)。
关键词
放顶煤
煤矸识别
听觉模型
听觉神经递质
特征融合
卷积神经网络
top coal caving
coal gangue recognition
auditory model
auditory neurotransmitter
feature fusion
convolutional neural network
