Transient scaling extraction chirplet transform: a time–frequency analysis tool for pulse feature extraction

Abstract Time–frequency analysis (TFA) technology is an effective tool for revealing hidden transient impulse features in signals. Current TFA methods are modeled based on the continuity assumption of time-domain signals, leading to issues such as time–frequency energy diffusion, inapplicability to impulse components, and over-squeezing of time–frequency features when extracting impulse characteristics. To achieve effective extraction of transient impulses, this paper develops a TFA method called transient scaling extraction CT (TSECT) with scaling-basis chirplet transform as its core. By using the Dirac function, which possesses time-localization capability, as the signal model, the time–frequency energy distribution of the transient structure of impulse components is deeply analyzed with respect to the scale basis function. This allows the construction of a transient scaling extraction operator to reassign the time–frequency coefficients and eliminate diffused time–frequency energy. Comparative analysis of simulated signals with impulse components demonstrates the effectiveness of TSECT. The generalization applicability of TSECT in the fields of mechanical fault diagnosis and biological signals is illustrated through vibration signals from real faulty bearings and echolocation signals of brown bats. Furthermore, while ensuring the complete characterization of time–frequency features, Rényi entropy (RE) and Stankovic concentration measure (SCM) are adopted as quantitative measures of time–frequency energy concentration. The numerical results of RE and SCM for TSECT are lower than those of other TFA methods, indicating that TSECT achieves superior energy concentration. Additionally, a comparison of computation times shows that the computational cost of TSECT is comparable to that of many advanced TFA methods.