Woosung Choi, Yeong-Seok Jeong, Jinsung Kim, Jaehwa Chung, Soonyoung Jung, J. Reiss
{"title":"通过自调节集中频率变换实现条件源分离","authors":"Woosung Choi, Yeong-Seok Jeong, Jinsung Kim, Jaehwa Chung, Soonyoung Jung, J. Reiss","doi":"10.17743/jaes.2022.0030","DOIUrl":null,"url":null,"abstract":"Label-conditioned source separation extracts the target source, specified by an input symbol, from an input mixture track. A recently proposed label-conditioned source separation model called Latent Source Attentive Frequency Transformation (LaSAFT)–Gated Point-Wise Con- volutional Modulation (GPoCM)–Net introduced a block for latent source analysis called LaSAFT. Employing LaSAFT blocks, it established state-of-the-art performance on several tasks of the MUSDB18 benchmark. This paper enhances the LaSAFT block by exploiting a self-conditioning method. Whereas the existing method only cares about the symbolic re- lationships between the target source symbol and latent sources, ignoring audio content, the new approach also considers audio content. The enhanced block computes the attention mask conditioning on the label and the input audio feature map. Here, it is shown that the conditioned U-Net employing the enhanced LaSAFT blocks outperforms the previous model. It is also shown that the present model performs the audio-query–based separation with a slight modification.","PeriodicalId":50008,"journal":{"name":"Journal of the Audio Engineering Society","volume":" ","pages":""},"PeriodicalIF":1.1000,"publicationDate":"2022-11-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Conditioned Source Separation by Attentively Aggregating Frequency Transformations With Self-Conditioning\",\"authors\":\"Woosung Choi, Yeong-Seok Jeong, Jinsung Kim, Jaehwa Chung, Soonyoung Jung, J. Reiss\",\"doi\":\"10.17743/jaes.2022.0030\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Label-conditioned source separation extracts the target source, specified by an input symbol, from an input mixture track. A recently proposed label-conditioned source separation model called Latent Source Attentive Frequency Transformation (LaSAFT)–Gated Point-Wise Con- volutional Modulation (GPoCM)–Net introduced a block for latent source analysis called LaSAFT. Employing LaSAFT blocks, it established state-of-the-art performance on several tasks of the MUSDB18 benchmark. This paper enhances the LaSAFT block by exploiting a self-conditioning method. Whereas the existing method only cares about the symbolic re- lationships between the target source symbol and latent sources, ignoring audio content, the new approach also considers audio content. The enhanced block computes the attention mask conditioning on the label and the input audio feature map. Here, it is shown that the conditioned U-Net employing the enhanced LaSAFT blocks outperforms the previous model. It is also shown that the present model performs the audio-query–based separation with a slight modification.\",\"PeriodicalId\":50008,\"journal\":{\"name\":\"Journal of the Audio Engineering Society\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":1.1000,\"publicationDate\":\"2022-11-02\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of the Audio Engineering Society\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://doi.org/10.17743/jaes.2022.0030\",\"RegionNum\":4,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"ACOUSTICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of the Audio Engineering Society","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.17743/jaes.2022.0030","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ACOUSTICS","Score":null,"Total":0}
Conditioned Source Separation by Attentively Aggregating Frequency Transformations With Self-Conditioning
Label-conditioned source separation extracts the target source, specified by an input symbol, from an input mixture track. A recently proposed label-conditioned source separation model called Latent Source Attentive Frequency Transformation (LaSAFT)–Gated Point-Wise Con- volutional Modulation (GPoCM)–Net introduced a block for latent source analysis called LaSAFT. Employing LaSAFT blocks, it established state-of-the-art performance on several tasks of the MUSDB18 benchmark. This paper enhances the LaSAFT block by exploiting a self-conditioning method. Whereas the existing method only cares about the symbolic re- lationships between the target source symbol and latent sources, ignoring audio content, the new approach also considers audio content. The enhanced block computes the attention mask conditioning on the label and the input audio feature map. Here, it is shown that the conditioned U-Net employing the enhanced LaSAFT blocks outperforms the previous model. It is also shown that the present model performs the audio-query–based separation with a slight modification.
期刊介绍:
The Journal of the Audio Engineering Society — the official publication of the AES — is the only peer-reviewed journal devoted exclusively to audio technology. Published 10 times each year, it is available to all AES members and subscribers.
The Journal contains state-of-the-art technical papers and engineering reports; feature articles covering timely topics; pre and post reports of AES conventions and other society activities; news from AES sections around the world; Standards and Education Committee work; membership news, patents, new products, and newsworthy developments in the field of audio.