| 基于深度学习的声信号分类识别方法 |
| |
| 引用本文： | 王鹏程,崔敏,王彦博,李剑,赵欣.基于深度学习的声信号分类识别方法[J].单片机与嵌入式系统应用,2021,21(1):23-26. |
| |
| 作者姓名： | 王鹏程 崔敏 王彦博 李剑 赵欣 |
| |
| 作者单位： | 中北大学信息探测与处理山西省重点实验室,太原030051;中北大学信息探测与处理山西省重点实验室,太原030051;中北大学信息探测与处理山西省重点实验室,太原030051;中北大学信息探测与处理山西省重点实验室,太原030051;北方科技信息研究所 |
| |
| 基金项目： | 装备预研兵器工业联合基金;国家自然基金青年科学基金;装备预研兵器装备联合基金;山西省高校创新项目;山西省高等学校科技成果转换培育项目;山西省面上青年资金;中北大学科学研究基金 |
| |
| 摘 要： | 提出了一种基于深度学习的声信号分类识别方法,将声场环境中声源目标的识别等效为声场信号—特定声源的端到端学习过程,建立一种以log-mel能量为声信号特征的预提取方法,以深度残差网络作为特征自动提取及分类的声信号分类识别模型。在两个大型数据集上对模型性能进行了验证,实验结果表明,本文提出的深度残差网络模型在DCASE2019数据集和UrbanSound8K数据集上能够实现80.2%和76.4%的识别精度,在声源探测领域具有一定的应用价值。
|
| 关 键 词： | 声源探测 声信号分类识别 深度学习 深度残差网络 时频域分析 |
Acoustic Signal Classification and Recognition Method Based on Deep Learning |
| |
| Wang Pengcheng,Cui Min,Wang Yanbo,Li Jian,Zhao Xin.Acoustic Signal Classification and Recognition Method Based on Deep Learning[J].Microcontrollers & Embedded Systems,2021,21(1):23-26. |
| |
| Authors: | Wang Pengcheng Cui Min Wang Yanbo Li Jian Zhao Xin |
| |
| Affiliation: | (Shanxi Province Key Laboratory of Information Detection and Processing,North University of China,Taiyuan 030051,China;North Institute of Science and Technology Information) |
| |
| Abstract: | In the paper,a deep learning-based sound signal classification and recognition method is proposed.The sound source target recognition in the sound field environment is equivalent to the sound field signal-end-to-end of a specific sound source.During the learning process,a sound signal classification and recognition model with log-mel energy as the pre-extraction method of acoustic signal features and deep residual network as the feature automatic extraction and classification is established.The performance of the model is verified on two large data sets.The experiment results show that the deep residual network model proposed in this paper can achieve 80.2%and 76.4%recognition accuracy on the DCASE2019 data set and UrbanSound8K data set.The detection field has certain application value. |
| |
| Keywords: | sound source detection acoustic signal classification and recognition deep learning deep residual network time-frequency domain analysis |
| 本文献已被 维普 万方数据 等数据库收录！ |
|
