用于轻量级声学场景分类的深空可分离蒸馏技术

ShuQi Ye, Yuan Tian
{"title":"用于轻量级声学场景分类的深空可分离蒸馏技术","authors":"ShuQi Ye, Yuan Tian","doi":"arxiv-2405.03567","DOIUrl":null,"url":null,"abstract":"Acoustic scene classification (ASC) is highly important in the real world.\nRecently, deep learning-based methods have been widely employed for acoustic\nscene classification. However, these methods are currently not lightweight\nenough as well as their performance is not satisfactory. To solve these\nproblems, we propose a deep space separable distillation network. Firstly, the\nnetwork performs high-low frequency decomposition on the log-mel spectrogram,\nsignificantly reducing computational complexity while maintaining model\nperformance. Secondly, we specially design three lightweight operators for ASC,\nincluding Separable Convolution (SC), Orthonormal Separable Convolution (OSC),\nand Separable Partial Convolution (SPC). These operators exhibit highly\nefficient feature extraction capabilities in acoustic scene classification\ntasks. The experimental results demonstrate that the proposed method achieves a\nperformance gain of 9.8% compared to the currently popular deep learning\nmethods, while also having smaller parameter count and computational\ncomplexity.","PeriodicalId":501178,"journal":{"name":"arXiv - CS - Sound","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Deep Space Separable Distillation for Lightweight Acoustic Scene Classification\",\"authors\":\"ShuQi Ye, Yuan Tian\",\"doi\":\"arxiv-2405.03567\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Acoustic scene classification (ASC) is highly important in the real world.\\nRecently, deep learning-based methods have been widely employed for acoustic\\nscene classification. However, these methods are currently not lightweight\\nenough as well as their performance is not satisfactory. To solve these\\nproblems, we propose a deep space separable distillation network. Firstly, the\\nnetwork performs high-low frequency decomposition on the log-mel spectrogram,\\nsignificantly reducing computational complexity while maintaining model\\nperformance. Secondly, we specially design three lightweight operators for ASC,\\nincluding Separable Convolution (SC), Orthonormal Separable Convolution (OSC),\\nand Separable Partial Convolution (SPC). These operators exhibit highly\\nefficient feature extraction capabilities in acoustic scene classification\\ntasks. The experimental results demonstrate that the proposed method achieves a\\nperformance gain of 9.8% compared to the currently popular deep learning\\nmethods, while also having smaller parameter count and computational\\ncomplexity.\",\"PeriodicalId\":501178,\"journal\":{\"name\":\"arXiv - CS - Sound\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-05-06\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Sound\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2405.03567\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Sound","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2405.03567","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

声学场景分类(ASC)在现实世界中非常重要。最近,基于深度学习的方法被广泛用于声学场景分类。然而,这些方法目前还不够轻便,性能也不尽如人意。为了解决这些问题,我们提出了一种深度空间可分离蒸馏网络。首先,该网络对log-mel频谱图进行高低频分解,在保持模型性能的同时大大降低了计算复杂度。其次,我们专门为 ASC 设计了三个轻量级算子,包括可分离卷积(SC)、正交可分离卷积(OSC)和可分离部分卷积(SPC)。这些算子在声学场景分类任务中表现出高效的特征提取能力。实验结果表明,与目前流行的深度学习方法相比,所提出的方法实现了 9.8% 的性能增益,同时还具有更少的参数数量和计算复杂性。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Deep Space Separable Distillation for Lightweight Acoustic Scene Classification
Acoustic scene classification (ASC) is highly important in the real world. Recently, deep learning-based methods have been widely employed for acoustic scene classification. However, these methods are currently not lightweight enough as well as their performance is not satisfactory. To solve these problems, we propose a deep space separable distillation network. Firstly, the network performs high-low frequency decomposition on the log-mel spectrogram, significantly reducing computational complexity while maintaining model performance. Secondly, we specially design three lightweight operators for ASC, including Separable Convolution (SC), Orthonormal Separable Convolution (OSC), and Separable Partial Convolution (SPC). These operators exhibit highly efficient feature extraction capabilities in acoustic scene classification tasks. The experimental results demonstrate that the proposed method achieves a performance gain of 9.8% compared to the currently popular deep learning methods, while also having smaller parameter count and computational complexity.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信