Symbolic time series analysis via wavelet-based partitioning

Symbolic time series analysis (STSA) of complex systems for anomaly detection has been recently introduced in literature. An important feature of the STSA method is extraction of relevant information, imbedded in the measured time series data, to generate symbol sequences. This paper presents a wavelet-based partitioning approach for symbol generation, instead of the currently practiced method of phase-space partitioning. Various aspects of the proposed technique, such as wavelet selection, noise mitigation, and robustness to spurious disturbances, are discussed. The wavelet-based partitioning in STSA is experimentally validated on laboratory apparatuses for anomaly/damage detection. Its efficacy is investigated by comparison with phase-space partitioning.