{"id":"https://openalex.org/W4372260014","doi":"https://doi.org/10.1109/icassp49357.2023.10095977","title":"Fixed-Point Quantization Aware Training for on-Device Keyword-Spotting","display_name":"Fixed-Point Quantization Aware Training for on-Device Keyword-Spotting","publication_year":2023,"publication_date":"2023-05-05","ids":{"openalex":"https://openalex.org/W4372260014","doi":"https://doi.org/10.1109/icassp49357.2023.10095977"},"language":"en","primary_location":{"id":"doi:10.1109/icassp49357.2023.10095977","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49357.2023.10095977","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5006855552","display_name":"Sashank Macha","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Sashank Macha","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5053087192","display_name":"Om Oza","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Om Oza","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068300265","display_name":"Alex Escott","orcid":"https://orcid.org/0000-0002-2683-624X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alex Escott","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014387832","display_name":"Francesco Caliv\u00e1","orcid":"https://orcid.org/0000-0002-0425-7511"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Francesco Caliv\u00e1","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013861601","display_name":"Robbie Armitano","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Robbie Armitano","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005677843","display_name":"Santosh Kumar Cheekatmalla","orcid":"https://orcid.org/0000-0003-1144-4507"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Santosh Kumar Cheekatmalla","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5080341961","display_name":"Sree Hari Krishnan Parthasarathi","orcid":null},"institutions":[{"id":"https://openalex.org/I1311688040","display_name":"Amazon (United States)","ror":"https://ror.org/04mv4n011","country_code":"US","type":"company","lineage":["https://openalex.org/I1311688040"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Sree Hari Krishnan Parthasarathi","raw_affiliation_strings":["Alexa AI,Amazon","Alexa AI, Amazon"],"affiliations":[{"raw_affiliation_string":"Alexa AI,Amazon","institution_ids":["https://openalex.org/I1311688040"]},{"raw_affiliation_string":"Alexa AI, Amazon","institution_ids":["https://openalex.org/I1311688040"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5012048094","display_name":"Yuzong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yuzong Liu","raw_affiliation_strings":["Alexa Perceptual Technologies"],"affiliations":[{"raw_affiliation_string":"Alexa Perceptual Technologies","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":8,"corresponding_author_ids":["https://openalex.org/A5006855552"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.5697,"has_fulltext":false,"cited_by_count":8,"citation_normalized_percentile":{"value":0.83464226,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"5"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11309","display_name":"Music and Audio Processing","score":0.9972000122070312,"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/T11309","display_name":"Music and Audio Processing","score":0.9972000122070312,"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/T11447","display_name":"Blind Source Separation Techniques","score":0.9965000152587891,"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/T12611","display_name":"Neural Networks and Reservoir Computing","score":0.9940000176429749,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.8439455032348633},{"id":"https://openalex.org/keywords/keyword-spotting","display_name":"Keyword spotting","score":0.7551113367080688},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6528064012527466},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5637342929840088},{"id":"https://openalex.org/keywords/flops","display_name":"FLOPS","score":0.5438582897186279},{"id":"https://openalex.org/keywords/floating-point","display_name":"Floating point","score":0.4792380630970001},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.4318715035915375},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.4031822085380554},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.39204108715057373},{"id":"https://openalex.org/keywords/computer-engineering","display_name":"Computer engineering","score":0.3845069408416748},{"id":"https://openalex.org/keywords/algorithm","display_name":"Algorithm","score":0.3226531147956848},{"id":"https://openalex.org/keywords/parallel-computing","display_name":"Parallel computing","score":0.19215744733810425}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8439455032348633},{"id":"https://openalex.org/C2781213101","wikidata":"https://www.wikidata.org/wiki/Q6398558","display_name":"Keyword spotting","level":2,"score":0.7551113367080688},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6528064012527466},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5637342929840088},{"id":"https://openalex.org/C3826847","wikidata":"https://www.wikidata.org/wiki/Q188768","display_name":"FLOPS","level":2,"score":0.5438582897186279},{"id":"https://openalex.org/C84211073","wikidata":"https://www.wikidata.org/wiki/Q117879","display_name":"Floating point","level":2,"score":0.4792380630970001},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.4318715035915375},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4031822085380554},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39204108715057373},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3845069408416748},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3226531147956848},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.19215744733810425}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/icassp49357.2023.10095977","is_oa":false,"landing_page_url":"https://doi.org/10.1109/icassp49357.2023.10095977","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":36,"referenced_works":["https://openalex.org/W1605005685","https://openalex.org/W1902934009","https://openalex.org/W2094756095","https://openalex.org/W2242818861","https://openalex.org/W2286365479","https://openalex.org/W2293968535","https://openalex.org/W2565125333","https://openalex.org/W2585560244","https://openalex.org/W2623451521","https://openalex.org/W2797583228","https://openalex.org/W2891335318","https://openalex.org/W2907106536","https://openalex.org/W2909067195","https://openalex.org/W2928062709","https://openalex.org/W2963122961","https://openalex.org/W2963917524","https://openalex.org/W2968917279","https://openalex.org/W2970601456","https://openalex.org/W2970714388","https://openalex.org/W2982344224","https://openalex.org/W3083008554","https://openalex.org/W3097522836","https://openalex.org/W3098576111","https://openalex.org/W3119699820","https://openalex.org/W3137841151","https://openalex.org/W3193844031","https://openalex.org/W4295937534","https://openalex.org/W4297841491","https://openalex.org/W6639703010","https://openalex.org/W6690026940","https://openalex.org/W6696004547","https://openalex.org/W6739513683","https://openalex.org/W6750665317","https://openalex.org/W6757067665","https://openalex.org/W6758145142","https://openalex.org/W6767164110"],"related_works":["https://openalex.org/W2963122961","https://openalex.org/W3160141544","https://openalex.org/W4287207127","https://openalex.org/W2803935332","https://openalex.org/W2777406049","https://openalex.org/W4297196174","https://openalex.org/W3099092507","https://openalex.org/W4296563896","https://openalex.org/W2971667683","https://openalex.org/W4310354755"],"abstract_inverted_index":{"Fixed-point":[0],"(FXP)":[1],"inference":[2,148],"has":[3,25],"proven":[4],"suitable":[5],"for":[6,73,79],"embedded":[7],"devices":[8],"with":[9,60,109],"limited":[10],"computational":[11],"resources,":[12],"and":[13,30,49,69,76,156,204,207],"yet":[14],"model":[15,74,104,189],"training":[16,24],"is":[17],"continually":[18],"performed":[19],"in":[20,112,131,201],"floating-point":[21],"(FLP).":[22],"FXP":[23,37,51],"not":[26],"been":[27],"fully":[28],"explored":[29],"the":[31,93,158],"non-trivial":[32],"conversion":[33],"from":[34],"FLP":[35,145],"to":[36,47,107,142,167],"presents":[38],"unavoidable":[39],"performance":[40,186],"drop.":[41],"We":[42,56,171],"propose":[43,77],"a":[44,128],"novel":[45,195],"method":[46],"train":[48],"obtain":[50],"convolutional":[52],"keyword-spotting":[53],"(KWS)":[54],"models.":[55,146,209],"combine":[57],"our":[58,123],"methodology":[59],"two":[61],"quantization-aware-training":[62],"(QAT)":[63],"techniques":[64,78],"\u2013":[65],"squashed":[66],"weight":[67],"distribution":[68],"absolute":[70],"cosine":[71],"regularization":[72],"parameters,":[75],"extending":[80],"QAT":[81],"over":[82],"transient":[83],"variables,":[84],"otherwise":[85],"neglected":[86],"by":[87,179],"previous":[88],"paradigms.":[89],"Experimental":[90],"results":[91],"on":[92,115],"Google":[94],"Speech":[95],"Commands":[96],"v2":[97],"dataset":[98],"show":[99,121],"that":[100,122,151,173,197],"we":[101,120,149,174],"can":[102,175],"reduce":[103,168,176],"precision":[105,144],"up":[106],"4-bit":[108],"no":[110],"loss":[111],"accuracy.":[113],"Furthermore,":[114],"an":[116],"in-house":[117],"KWS":[118,183],"dataset,":[119],"8bit":[124],"FXP-QAT":[125,152],"models":[126],"have":[127],"4-6%":[129],"improvement":[130],"relative":[132],"false":[133,138],"discovery":[134],"rate":[135,140],"at":[136],"fixed":[137],"reject":[139],"compared":[141],"full":[143],"During":[147],"argue":[150],"eliminates":[153],"q-format":[154],"normalization":[155],"enables":[157],"use":[159],"of":[160],"low-bit":[161],"accumulators":[162],"while":[163],"maximizing":[164],"SIMD":[165],"throughput":[166],"user-perceived":[169],"latency.":[170],"demonstrate":[172],"execution":[177],"time":[178],"68%":[180],"without":[181],"compromising":[182],"model\u2019s":[184],"predictive":[185],"or":[187],"requiring":[188],"architectural":[190],"changes.":[191],"Our":[192],"work":[193],"provides":[194],"findings":[196],"aid":[198],"future":[199],"research":[200],"this":[202],"area":[203],"enable":[205],"accurate":[206],"efficient":[208]},"counts_by_year":[{"year":2026,"cited_by_count":1},{"year":2025,"cited_by_count":2},{"year":2024,"cited_by_count":5}],"updated_date":"2026-04-08T06:01:36.053099","created_date":"2025-10-10T00:00:00"}