{"id":"https://openalex.org/W7106145428","doi":"https://doi.org/10.48550/arxiv.2511.14515","title":"IMSE: Efficient U-Net-based Speech Enhancement using Inception Depthwise Convolution and Amplitude-Aware Linear Attention","display_name":"IMSE: Efficient U-Net-based Speech Enhancement using Inception Depthwise Convolution and Amplitude-Aware Linear Attention","publication_year":2025,"publication_date":"2025-11-18","ids":{"openalex":"https://openalex.org/W7106145428","doi":"https://doi.org/10.48550/arxiv.2511.14515"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2511.14515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2511.14515","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2511.14515","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Tang, Xinxin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Xinxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Qin, Bin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qin, Bin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Li, Yufang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Yufang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.9904000163078308,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10860","display_name":"Speech and Audio Processing","score":0.9904000163078308,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10201","display_name":"Speech Recognition and Synthesis","score":0.0038999998942017555,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11233","display_name":"Advanced Adaptive Filtering Techniques","score":0.0010000000474974513,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/pesq","display_name":"PESQ","score":0.7089999914169312},{"id":"https://openalex.org/keywords/speech-enhancement","display_name":"Speech enhancement","score":0.66839998960495},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.5764999985694885},{"id":"https://openalex.org/keywords/convolution","display_name":"Convolution (computer science)","score":0.5548999905586243},{"id":"https://openalex.org/keywords/spectrogram","display_name":"Spectrogram","score":0.5178999900817871},{"id":"https://openalex.org/keywords/metric","display_name":"Metric (unit)","score":0.4503999948501587},{"id":"https://openalex.org/keywords/offset","display_name":"Offset (computer science)","score":0.4487999975681305},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.4487000107765198}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7275999784469604},{"id":"https://openalex.org/C103734657","wikidata":"https://www.wikidata.org/wiki/Q2739975","display_name":"PESQ","level":4,"score":0.7089999914169312},{"id":"https://openalex.org/C2776182073","wikidata":"https://www.wikidata.org/wiki/Q7575395","display_name":"Speech enhancement","level":3,"score":0.66839998960495},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.5764999985694885},{"id":"https://openalex.org/C45347329","wikidata":"https://www.wikidata.org/wiki/Q5166604","display_name":"Convolution (computer science)","level":3,"score":0.5548999905586243},{"id":"https://openalex.org/C45273575","wikidata":"https://www.wikidata.org/wiki/Q578970","display_name":"Spectrogram","level":2,"score":0.5178999900817871},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.4964999854564667},{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.4503999948501587},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.4487999975681305},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.4487000107765198},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.44699999690055847},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3882000148296356},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.3626999855041504},{"id":"https://openalex.org/C61328038","wikidata":"https://www.wikidata.org/wiki/Q3358061","display_name":"Speech processing","level":2,"score":0.3336000144481659},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.325300008058548},{"id":"https://openalex.org/C131109320","wikidata":"https://www.wikidata.org/wiki/Q581012","display_name":"Linear prediction","level":2,"score":0.31700000166893005},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.29580000042915344},{"id":"https://openalex.org/C191795146","wikidata":"https://www.wikidata.org/wiki/Q3878446","display_name":"Norm (philosophy)","level":2,"score":0.2888000011444092},{"id":"https://openalex.org/C163175372","wikidata":"https://www.wikidata.org/wiki/Q3339222","display_name":"Linear model","level":2,"score":0.258899986743927},{"id":"https://openalex.org/C204201278","wikidata":"https://www.wikidata.org/wiki/Q1332614","display_name":"Voice activity detection","level":3,"score":0.2524000108242035},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.251800000667572},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.2506999969482422}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2511.14515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2511.14515","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2511.14515","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2511.14515","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Achieving":[0],"a":[1,10,29,38,65,91,220],"balance":[2],"between":[3,226],"lightweight":[4],"design":[5],"and":[6,44,94,167,229],"high":[7],"performance":[8,207],"remains":[9],"significant":[11],"challenge":[12],"for":[13,80,223],"speech":[14,230,234],"enhancement":[15],"(SE)":[16],"tasks":[17],"on":[18,64,181,212],"resource-constrained":[19],"devices.":[20],"Existing":[21],"state-of-the-art":[22,211],"methods,":[23],"such":[24],"as":[25],"MUSE,":[26],"have":[27],"established":[28],"strong":[30],"baseline":[31],"with":[32,107,148,174],"only":[33],"0.51M":[34],"parameters":[35],"by":[36,121,198],"introducing":[37],"Multi-path":[39],"Enhanced":[40],"Taylor":[41],"(MET)":[42],"transformer":[43],"Deformable":[45],"Embedding":[46],"(DE).":[47],"However,":[48],"an":[49,139],"in-depth":[50],"analysis":[51],"reveals":[52],"that":[53],"MUSE":[54,190],"still":[55],"suffers":[56],"from":[57],"efficiency":[58],"bottlenecks:":[59],"the":[60,71,77,104,115,124,131,145,155,182,189,195,210,213,224],"MET":[61,105],"module":[62,106,147],"relies":[63],"complex":[66],"\"approximate-compensate\"":[67],"mechanism":[68],"to":[69,188,202,209],"mitigate":[70],"limitations":[72],"of":[73,127],"Taylor-expansion-based":[74],"attention,":[75],"while":[76,204],"offset":[78],"calculation":[79],"deformable":[81],"embedding":[82],"introduces":[83],"additional":[84],"computational":[85],"burden.":[86],"This":[87,217],"paper":[88],"proposes":[89],"IMSE,":[90],"systematically":[92],"optimized":[93],"ultra-lightweight":[95,233],"network.":[96],"We":[97],"introduce":[98],"two":[99],"core":[100],"innovations:":[101],"1)":[102],"Replacing":[103,144],"Amplitude-Aware":[108],"Linear":[109],"Attention":[110],"(MALA).":[111],"MALA":[112],"fundamentally":[113],"rectifies":[114],"\"amplitude-ignoring\"":[116],"problem":[117],"in":[118,130,232],"linear":[119],"attention":[120,132],"explicitly":[122],"preserving":[123],"norm":[125],"information":[126],"query":[128],"vectors":[129],"calculation,":[133],"achieving":[134,205],"efficient":[135,162],"global":[136],"modeling":[137],"without":[138],"auxiliary":[140],"compensation":[141],"branch.":[142],"2)":[143],"DE":[146],"Inception":[149,156],"Depthwise":[150],"Convolution":[151],"(IDConv).":[152],"IDConv":[153],"borrows":[154],"concept,":[157],"decomposing":[158],"large-kernel":[159],"operations":[160],"into":[161],"parallel":[163],"branches":[164],"(square,":[165],"horizontal,":[166],"vertical":[168],"strips),":[169],"thereby":[170],"capturing":[171],"spectrogram":[172],"features":[173],"extremely":[175],"low":[176],"parameter":[177,196],"redundancy.":[178],"Extensive":[179],"experiments":[180],"VoiceBank+DEMAND":[183],"dataset":[184],"demonstrate":[185],"that,":[186],"compared":[187],"baseline,":[191],"IMSE":[192],"significantly":[193],"reduces":[194],"count":[197],"16.8\\%":[199],"(from":[200],"0.513M":[201],"0.427M)":[203],"competitive":[206],"comparable":[208],"PESQ":[214],"metric":[215],"(3.373).":[216],"study":[218],"sets":[219],"new":[221],"benchmark":[222],"trade-off":[225],"model":[227],"size":[228],"quality":[231],"enhancement.":[235]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2025-11-20T00:00:00"}