A State Machine-Based Dual-Threshold Heart Sound Segmentation Method

Early detection of cardiovascular pathologies relies fundamentally on precise segmentation of phonocardiogram (PCG) signals, particularly for identifying valvular dysfunction and cardiac decompensation. We propose a finite state automaton-coordinated dual-threshold framework to resolve three persistent challenges in phonocardiogram signal analysis: signal instability, ambient noise contamination, and rhythm variability. The technical workflow comprises: 1) Morphological enhancement through Viola's envelope extraction, which suppresses high-frequency artifacts while preserving pathological signatures; 2) State-transition controlled threshold adaptation generating dynamic TGS markers for cardiac event localization (S1/S2 boundaries and hemodynamic phases); 3) Context-aware segment fusion compensating for acquisition gaps. Comparative evaluations against the Springer benchmark reveal a 96.45% reduction in computational latency (3.55% residual time) with 82% segmentation fidelity across 600 clinical samples. The architecture's parameter-tunable design circumvents data-intensive training requirements, achieving diagnostic-grade precision with 58% fewer computational operations than conventional methods. Clinically, this low-resource solution demonstrates deployability in primary care units through its hardware-agnostic implementation and traceable decision logic.