Perceptual Hashing of Cyclostationary Signal with Sparse Coding
Haining Liu, Yixiang Huang, Chengliang Liu, Jinkai Zhang · 2018
Cyclostationarity is the most common property displayed in the monitored signals during a machine condition monitoring process, especially for rotating and reciprocating machines. Redundancy in cyclostationary signals is also obvious, not only because these signals are usually sampled with high frequency, but also the machine health information remains the same and stable for a period of time. Thus, the transmission, storage and computation with these signals would unnecessarily consume more resources, power and costs. In order to reduce the redundancy, optimize the condition monitoring process, a new framework is introduced based on perceptual hashing, and sparse coding is proposed as an effective method to generate machine condition hash (MCH). The verification with bearing vibration signals shows that the data dimensionality can be reduced while diagnostic information can be retrieved. Finally, bandwidth consumption and data storage space can be reduced, while the diagnostic accuracy can still be guaranteed.