To improve the accuracy of identifying failure precursors in water?soaked coal, uniaxial compression and synchronous acoustic emission?(AE) experiments were performed on coal specimens in the natural state and after 120?h of water immersion, addressing the problem that a high proportion of low-energy AE events masks high-energy information during fracturing. An AE signal screening method coupling energy contribution ratio with damage stress was established to extract dominant signals and analyse their precursory characteristics. Results show that (1) water immersion markedly reduces coal compressive strength, promotes multidirectional crack propagation, and intensifies ultimate failure; (2) AE count and cumulative AE count during fracturing serve as key parameters for recognising impending failure, but their reliability must be judged in conjunction with specific fracturing stages; (3) the proposed method efficiently extracts dominant signals during the fracture of soaked coal, yielding lower energy thresholds and higher dominant-signal proportions than in natural coal; (4) the extracted dominant signals correspond closely to the precursory characteristics of soaked?coal failure and provide an effective basis for engineering stability prediction.