@article{4807,
  author = {Jie-Wei Name: Zhang},
  title = {Robust Change-Point Detection of Tactical Engagement Transitions in Underwater Telemetry Using Multimethod Statistical Analysis},
  journal = {Progress in Signals and Telecommunication Engineering},
  year = {2026},
  volume = {15},
  number = {2},
  doi = {https://doi.org/10.6025/pste/2026/15/2/49-63},
  url = {https://www.dline.info/pste/fulltext/v15n2/pstev15n2_1.pdf},
  abstract = {Reliable identification of tactical engagement transitions in continuous underwater telemetry is critical for
automated monitoring and situational awareness, yet it remains challenging due to severe bandwidth constraints,
environmental noise, and complex multivariate sensor dynamics. This study develops and evaluates a
robust multimethod change point detection framework designed to accurately identify these operational
transitions. We applied seven complementary statistical approaches, CUSUM, PELT, binary segmentation,
Bayesian inference, kernel detection, ruptures style segmentation, and multivariate Mahalanobis analysis
to a comprehensive 12,000-observation telemetry dataset. Dimensionality reduction via principal component
analysis and multivariate distributional metrics were utilized to capture coordinated, system level state
shifts.
Results demonstrated exceptional cross-method convergence and temporal precision. Six of the seven methods
localized the tactical engagement onset exactly at the reference boundary, while the online CUSUM monitor
detected the transition within a negligible five-second delay. Furthermore, segmentation methods
successfully recovered the complete engagement interval, accurately identifying both onset and termination
boundaries. Multivariate and kernel analyses confirmed that the transition represents a profound, systemwide
distributional shift rather than an isolated sensor anomaly.
By integrating diverse detection principles, the proposed framework provides strong cross validation,
ensuring the identified boundaries are genuine features of the underlying telemetry dynamics rather than
algorithmic artifacts. Ultimately, this multimethod approach offers a highly reliable, automated mechanism
for distinguishing routine patrol from event related tactical states, significantly enhancing real time decisionmaking,
reducing operator workload, and improving mission reliability in complex underwater sensing
environments.},
}