With　the　increasingly　severe　global　climate　situation,　the　intensifying　urban　heat　island　effect,　and　the　widespread　impact　of　COVID-19,　people＇s　expectations　for　a　healthy　and　comfortable　indoor　environment　continue　to　rise　along　with　the　improvement　of　their　quality　of　life.　Traditional　heating,　ventilation,　and　air　conditioning　(HVAC)　systems　primarily　provide　comfort　by　setting　a　fixed　temperature　point.　This　passive　adjustment　method　not　only　fails　to　dynamically　adapt　to　the　differentiated　needs　of　various　household　compositions　but　also　results　in　energy　wastage.　As　a　result,　this　study　proposes　a　user-centric　indoor　comfort　adjustment　strategy　that　provides　refined　temperature　control　for　different　households,　integral　to　the　concept　of　smart　homes.　Considering　the　impact　of　user　characteristics　(such　as　age　or　health　status)　and　activity　levels　(such　as　sitting　or　exercising)　on　comfort,　the　Analytic　Hierarchy　Process　(AHP)　is　used　to　allocate　weights　to　collected　human　physiological　data.　Subsequently,　the　Human　Comfort　Bi-directional　Long　Short-Term　Memory　(HC-BiLSTM)　network　is　employed　to　predict　comfort　levels　more　accurately.　The　predictions　are　then　converted　into　real-time　control　commands　for　indoor　smart　home　devices　through　reinforcement　learning　techniques,　achieving　a　more　dynamic　and　refined　indoor　environment　adjustment.　This　temperature　control　strategy　is　centered　on　the　user,　providing　personalized　temperature　control　for　individuals　and　optimal　temperature　strategies　for　multiple　occupants　through　proactive,　unobtrusive　interaction,　thus　enhancing　the　home　living　experience.　The　adoption　of　this　strategy　offers　new　solutions　for　refined　services　in　smart　homes　and　energy　optimization,　contributing　to　the　advancement　of　smart　home　technologies　towards　more　intelligent　and　personalized　directions.　©　2024　IEEE.