[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128577-en":3,"doc-seo-128577-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},128577,549768064778,"Finn","https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8",8,"Research & Report","Inference and Denoise - Causal Inference-Based Neural Speech Enhancement","This study addresses the speech enhancement (SE) task under a causal inference paradigm by treating noise presence as an intervention. Using the potential outcome framework, the proposed causal inference-based speech enhancement model separates clean and noisy frames and routes each set through two mask-based enhancement modules guided by a noise detector. Training leverages noise presence for module selection, while inference selects modules per frame based on predicted noise presence. A SE-specific average treatment effect is derived to quantify causal impact. Experiments show improved performance and efficiency versus non-causal mask-based baselines and more complex SE models.","INFERENCE AND DENOISE: CAUSAL INFERENCE-BASED NEURAL SPEECH ENHANCEMENT  \nTsun-An Hsieh 1, Chao-Han Huck Yang2, Pin-Yu Chen3 , Sabato Marco Siniscalchi 2 ;4, Yu Tsao 1  \n1Research Center for Information Technology Innovation, Academia Sinica, Taiwan  \n2 Georgia Institute of Technology, GA, USA; 3IBM Research, NY, USA  \n4 Computer Engineering School, Norwegian University of Science and Technology, Norway  \narXiv :2211 .01189v1 [ ee ss .AS] 2 Nov 2022  \nABSTRACT  \nThis study addresses the speech enhancement (SE) task within the causal inference paradigm by modeling the noise presence as an intervention. Based on the potential outcome framework, the proposed causal inference-based speech enhancement (CISE) separates clean and noisy frames in an intervened noisy speech using a noise detector and assigns both sets of frames to two mask-based enhancement modules (EMs) to perform noise-conditional SE. Specifically, we use the presence of noise as guidance for EM selection during training, and the noise detector selects the enhancement module according to the prediction of the presence of noise for each frame. Moreover, we derived a SE-specific average treatment effect to quantify the causal effect adequately. Experimental evidence demonstrates that CISE outperforms a non-causal mask-based SE approach in the studied settings and has better performance and efficiency than more complex SE models.  \nIndex Terms— Observational Inference, Deep Causal Inference, and Speech Enhancement  \n1. INTRODUCTION  \nRecent advances in neural network-based speech enhancement (SE) have demonstrated impressive performance in terms of speech quality and intelligibility scores, such as perceptual evaluation of speech quality (PESQ) [1] and short-time objective intelligibility (STOI) [2] in various speech applications. However, modern SE approaches [3, 4, 5, 6, 7] do not explicitly take the presence of noise into account., and real-world acoustic scenarios often encounter inevitable observational uncertainties, For instance, a meeting could be abruptly disconnected, or a session could be disrupted by temporary noise from the external environment. That is, noise intervention may not affect the entire speech waveform. In such a scenario, conventional neural SE solutions may be unreliable for handling the intermittent/sporadic nature of the noise. By contrast, causal inference (CI) [8] may bea viable paradigm for performing SE. The design of an end-to-end neural SE model within the CI framework is the research question addressed in this study. Causal inference-based machine learning techniques are often featured with the ability to identify unobserved factors or features (also know as confounding variables) with improved model prediction and generalization, i.e., CI-based models are proven to be advantageous of tackling unseen data hence dependable [9, 10] . Furthermore, machine learning models that satisfy the CI training objectives stand to benefit from additional interpretable scores to formally quantify the causal effects, for example, in treatment effect estimation. Previous studies [11, 12] have demonstrated that learning to measure causal variables empowers effective  \n(a) Training  \n(b) Testing  \nFig. 1: Causal graphical model (CGM) for the training phase (a) and the testing phase (b) . The blue nodes x and y are observable (e.g., noisy speech and speech intelligibility scores) . Node z, colored white, is not observable as a parameterized latent variable. Node i, colored blue in (a), is only observable during training, and to be inferred by proposed causal model in (b) the testing time.  \nmodel selection. Meanwhile, similar designs are sparse expert models [13, 14] . These approaches divide a large task into small subtasks by allocating data categorized in different attributes to several local expert models. Nevertheless, those models do not take into account the assumption of a causal graph; therefore, they can not be evaluated under a formal causal learni","cbCaiuAWOZavqOuT","https://ap.wps.com/l/cbCaiuAWOZavqOuT","pdf",853480,1,6,"English","en",105,"# Abstract\n# Introduction\n# Background\n## Causal Inference & Represent","[{\"question\":\"How does the proposed model treat noise in the causal inference framework?\",\"answer\":\"Noise presence is modeled as an intervention rather than relying on direct correlation between noisy and clean speech. This guides how frames are processed during enhancement.\"},{\"question\":\"What role does the noise detector play during training and inference?\",\"answer\":\"During training, noise presence labels guide the selection of enhancement modules. During inference, the noise detector predicts noise presence for each frame and selects the corresponding enhancement module.\"},{\"question\":\"How is the causal effect of the intervention quantified in this work?\",\"answer\":\"The study derives a SE-specific average treatment effect to measure the causal influence of the selected noise-related intervention on enhancement outcomes.\"}]","Inference and Denoise - Causal Inference-Based Neural Speech Enhancement | PDF",1786001901,15,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"inference-and-denoise-causal-inference-based-neural-speech-enhancement","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/inference-and-denoise-causal-inference-based-neural-speech-enhancement/128577/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-23","2026-08-06",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"How does the proposed model treat noise in the causal inference framework?","Question",{"text":76,"@type":77},"Noise presence is modeled as an intervention rather than relying on direct correlation between noisy and clean speech. This guides how frames are processed during enhancement.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What role does the noise detector play during training and inference?",{"text":81,"@type":77},"During training, noise presence labels guide the selection of enhancement modules. During inference, the noise detector predicts noise presence for each frame and selects the corresponding enhancement module.",{"name":83,"@type":74,"acceptedAnswer":84},"How is the causal effect of the intervention quantified in this work?",{"text":85,"@type":77},"The study derives a SE-specific average treatment effect to measure the causal influence of the selected noise-related intervention on enhancement outcomes.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]