The　existing　porn　streamers　audio　recognition　algorithms　show　poor　performance　in　increasingly　complex　network　environment.　To　resolve　this　problem,　a　porn　streamer　audio　recognition　algorithm　based　on　deep　learning　and　random　forest　is　proposed.　In　this　algorithm,　a　more　stable　complementary　feature　is　first　proposed,　which　consists　of　Log　Mel　Spectrum　(LMS),　Mel　Frequency　Cepstrum　Coefficient　(MFCC)　and　Gammatone　Frequency　Cepstrum　Coefficient　(GFCC),　and　the　Dual-Path　Fused　Transformer　Net　(DPFTNet)　network　structure　is　then　proposed　for　sound　classification,　which　parallelizes　the　two　main　modules　of　the　Swin　Transformer,　so　that　more　feature　details　can　be　retained.　Finally,　the　random　forest　is　utilized　to　identify　porn　streamer.　The　experimental　results　show　that　this　algorithm　has　higher　recognition　accuracy　than　the　comparison　algorithm.　©　2023,　The　Author(s),　under　exclusive　licence　to　Springer　Science+Business　Media,　LLC,　part　of　Springer　Nature.