D-2-DenseNet噪音鲁棒的城市音频分类模型

doi:10.13190/j.jbupt.2020-122

北京邮电大学学报 ›› 2021, Vol. 44 ›› Issue (1): 86-91.doi: 10.13190/j.jbupt.2020-122

D-2-DenseNet噪音鲁棒的城市音频分类模型

曹毅, 黄子龙, 盛永健, 刘晨, 费鸿博

江南大学机械工程学院, 无锡 214122

收稿日期:2020-08-16 出版日期:2021-02-28 发布日期:2021-09-30
作者简介:曹毅(1974-),男,教授,硕士生导师,E-mail:caoyi@jiangnan.edu.cn.
基金资助:
国家自然科学基金项目（51375209）；江苏省"六大人才高峰"计划项目（ZBZZ-012）；高等学校学科创新引智计划项目（B18027）；江苏省研究生创新计划项目（JNKY20_1928）

Noise Robust Urban Audio Classification Based on 2-Order Dense Convolutional Network Using Dual Features

CAO Yi, HUANG Zi-long, SHENG Yong-jian, LIU Chen, FEI Hong-bo

School of Mechanical Engineering, Jiangnan University, Wuxi 214122, China

Received:2020-08-16 Online:2021-02-28 Published:2021-09-30

摘要/Abstract

摘要： 为了提高噪音环境下城市音频分类系统的鲁棒性，提出了一种双特征2阶密集卷积神经网络（D-2-DenseNet）噪音鲁棒的城市音频分类模型.首先介绍了噪音添加和噪音鲁棒处理，阐述了一种双特征互补偿的算法；然后结合2阶密集卷积神经网络与自适应机制提出了一种噪音鲁棒音频分类模型：双特征2阶密集卷积神经网络.模型采用双特征互补偿自适应算法，可在特征提取与模型训练中更有针对性地提取有效音频信息，降低噪音干扰，以提高噪音鲁棒性.最后，基于Dcase2016数据集开展噪音环境下城市音频分类测试.实验结果表明，模型分类准确率分别可达77.12%、75.52%，与基线模型相比，平均分类准确率分别提高了8.51%和10.38%，验证了模型良好的噪音鲁棒性.

关键词: 城市音频分类, 噪音鲁棒性, 双特征互补偿, 2阶密集卷积神经网络, 双特征2阶密集卷积神经网络

Abstract: A noise robust urban sound event classification model based on 2-order dense convolutional network using dual features(D-2-DenseNet) is proposed,which aims at the problems of insufficient robustness of current models. Firstly,the brief introduction of the method of noise adding and robust processing is presented. Moreover,a dual feature mutual compensation algorithm and 2-order dense convolutional network is presented. Meanwhile,a noise robust urban sound event classification model based on 2-DenseNet using dual features,i.e.D-2-DenseNet is proposed. Theoretically,D-2-DenseNet combines the advantages of feature compensation and 2-order dense convolutional neural network. The dual feature mutual compensation adaptive algorithm can effectively extract audio information and reduce noise interference to improve noise robustness. Finally,in order to validate advantages of the D-2-DenseNet,this new model is exploited in the urban sound event classification based on Dcase2016 datasets. Under conditions of channel noise and environmental noise,the experiment shows that the accuracy of the network is respectively 77.12% and 75.52%,which has added 8.51% and 10.38% compared with baseline. The noise robustness of D-2-DenseNet are also effectively verified.

Key words: sound event classification, noise robust, dual features mutual compensation, 2-order dense convolutional network, 2-order dense convolutional network using dual features

中图分类号:

TP391.42

曹毅, 黄子龙, 盛永健, 刘晨, 费鸿博. D-2-DenseNet噪音鲁棒的城市音频分类模型[J]. 北京邮电大学学报, 2021, 44(1): 86-91.

CAO Yi, HUANG Zi-long, SHENG Yong-jian, LIU Chen, FEI Hong-bo. Noise Robust Urban Audio Classification Based on 2-Order Dense Convolutional Network Using Dual Features[J]. Journal of Beijing University of Posts and Telecommunications, 2021, 44(1): 86-91.

参考文献

[1] 王静远, 李超, 熊璋, 等. 以数据为中心的智慧城市研究综述[J]. 计算机研究与发展, 2014, 51(2):239-259. Wang Jingyuan, Li Chao, Xiong Zhang, et al. Survey of data centric smart city[J]. Journal of Computer Research and Development, 2014, 51(2):239-259.
[2] Lim M, Lee D, Hosung P, et al. Convolutional neural network based on audio event classification[J]. KSII Transactions on Internet and Information Systems, 2018, 12:2748-2759.
[3] Zhang Xiaohu, Zou Yuexian, Shi Wei. Dilated convolution neural network with leaky ReLU for environmental sound classification acoustics[C]//22nd International Conference on Digital Signal Processing. Washington:IEEE Press, 2017:123-132.
[4] 蔡尚, 金鑫, 高盛翔, 等. 用于噪音鲁棒性语音识别的子带能量规整感知线性预测系数[J]. 声学学报, 2012, 37(6):667-672. Cai Shang, Jin Xin, Gao Shengxiang, et al. Sub-band power normalized perceptual linear predictive coefficient for robust automatic speech recognition[J]. ACTA Acustica, 2012, 37(6):667-672.
[5] Yu Tsao, Lin Tzu-Hao, Chen Fei, et al. Robust S1 and S2 heart sound recognition based on spectral restoration and multi-style training[J]. Biomedical Signal Processing and Control, 2019, 49:173-180.
[6] Ilyas O, Zeynep O, Oghuz F. Noise robust sound event classification with convolutional neural network[J]. Neurocomputing, 2018, 272:505-512.
[7] 易江燕, 陶建华, 刘斌, 等. 基于迁移学习的噪音鲁棒语音识别声学建模[J]. 清华大学学报, 2018, 58(1):55-60. Yi Jiangyan, Tao Jianhua, Liu Bin et al. Transfer learning for acoustic modeling of noise robust speech recognition[J]. Journal of Tsinghua University, 2018, 58(1):55-60.
[8] 仇景明, 曲桦, 赵季红. 一种鲁棒网络流量分类及新类型发现算法[J]. 北京邮电大学学报, 2020, 43(2):40-45. Qiu Jingming, Qu Hua, Zhao Jihong. A robust network traffic classification and new type discovery algorithm[J]. Journal of Beijing University of Posts and Telecommunications, 2020, 43(2):40-45.
[9] Johnson H.Noise data[EB/OL].United Kingdom Signal Processing Information Base, 1990[2020-08-16].http://spib.linse.ufsc.br/noise.html.
[10] Wang Dong,Wu Dalei,Zhu Xiaoyan.THCHS-30:a free Chinese speech corpus[EB/OL].Beijing:Tsinghua University,2015(2015-12-10)[2020-08-16]. http://arxiv.org/abs/1512.01882.2015.
[11] Sumithra M, Prakash D, Yuan X. Adaptive wavelet shrinkage for noise robust speaker recognition[J]. Digital Signal Processing, 2014, 33:180-190.
[12] 王雷, 王智广. 考虑多种特征因素的设计模式自动识别[J]. 北京邮电大学学报, 2017, 40(5):36-42. Wang Lei, Wang Zhiguang. Automatic design pattern detection on the consideration of multiple characteristic factors[J]. Journal of Beijing University of Posts and Telecommunications, 2017, 40(5):36-42.
[13] 曹毅, 黄子龙, 张威, 等. N-DenseNet的城市声音事件分类模型[J]. 西安电子科技大学学报, 2019, 46(6):9-16, 94. Cao Yi, Huang Zilong, Zhang Wei, et al. Urban sound event classification with n-order dense convolutional network[J]. Journal of Xidian University, 2019, 46(6):9-16, 94.
[14] Huang Gao, Liu Zhuang, Weinberger K Q. Densely connected convolutional networks[C]//30th IEEE Conference on Computer Vision and Pattern Recognition CVPR 2017. Honolulu:IEEE press, 2017:2261-2269.
[15] Huang Zilong, Liu Chen, Fei Hongbo, et al. Urban sound classification based on 2-order dense convolutional network using dual features[J/OL]. Applied Acoustics, 2020, 164[2020-08-16]. https://doi.org/10.1016/j.apacoust.2020.107243.
[16] Mesaros A, Heittola T, Virtanen T. TUT database for acoustic scene classification and sound event detection[C]//24th European Signal Processing Conference. Budapest:IEEE Press, 2016:3-9.

D-2-DenseNet噪音鲁棒的城市音频分类模型

Noise Robust Urban Audio Classification Based on 2-Order Dense Convolutional Network Using Dual Features

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价