Refined composite multiscale fluctuation-based dispersion Lempel-Ziv complexity for signal analysis

Area of Science:

• Signal Processing
• Time Series Analysis
• Acoustics

Background:

• Dispersion Lempel-Ziv Complexity (DLZC) and multiscale DLZC (MDLZC) are recent complexity indicators for acoustic time series.
• These methods, based on dispersion entropy, identify time series characteristics but lack fluctuation information and stability.

Purpose of the Study:

• To enhance time series complexity analysis by incorporating fluctuation information.
• To develop a more stable and informative complexity indicator for acoustic signals.
• To improve the performance of bearing fault diagnosis and ship signal classification.

Main Methods:

• Introduced Fluctuation-based DLZC (FDLZC) by adding fluctuation information to DLZC.
• Developed Refined Composite Multiscale FDLZC (RCMFDLZC) with an improved coarse-graining operation for enhanced stability and feature extraction.
• Applied the minimum redundancy maximum relevance (mRMR) algorithm for feature selection.

Main Results:

• RCMFDLZC features demonstrated superior separability and clustering effects in bearing fault and ship radiated noise signals.
• The RCMFDLZC-based signal analysis achieved higher recognition rates in bearing fault diagnosis and ship signal classification compared to other methods.

Conclusions:

• FDLZC and RCMFDLZC offer improved time series complexity quantification by including fluctuation information and enhancing stability.
• RCMFDLZC provides a robust feature extraction method for acoustic signal analysis, leading to more accurate fault diagnosis and classification.