A Dual-Channel Signal Quality Assessment Framework for Reliable Seated Ballistocardiographic Biosensing with Measurement Configuration Optimization

Ballistocardiography (BCG) enables unobtrusive physiological monitoring, but its signal quality is sensitive to body–sensor coupling and posture. This study proposes a dual-channel quality assessment framework for seated BCG and measurement-configuration evaluation. A chair-integrated Mach–Zehnder interferometer fiber-optic sensing system recorded back and hip BCG signals from 11 participants at five backrest angles (100°, 120°, 140°, 160°, and 180°). Thirty candidate features were extracted from 15 s windows, and channel-specific three-class quality models were developed using selected interpretable features and a bagged-trees classifier under repeated subject-wise evaluation. Classification probabilities were mapped to continuous quality scores and integrated with JJ-interval consistency and cross-channel delay stability to construct a dual-channel quality index. The models distinguished low-, medium-, and high-quality windows, while the joint index characterized differences in dual-channel analyzability across measurement configurations. Angle-related differences were not statistically significant when all windows were included but became significant after excluding low-quality windows. Within the tested cohort and experimental configuration, 140∘ showed the highest descriptive overall quality, followed by 160∘, whereas 100∘ performed worst. The angle ranking was largely stable across alternative weighting schemes. However, cross-channel consistency does not constitute independent physiological validation, and improvements in downstream parameter-estimation accuracy remain unverified. The proposed framework supports quality-guided evaluation of seated BCG acquisition configurations and warrants further validation against independent physiological references.