A New Regularized Adaptive Windowed Lomb Periodogram for Time–Frequency Analysis of Nonstationary Signals With Impulsive Components

摘要: This paper proposes a new class of windowed Lomb periodogram (WLP) for time-frequency analysis nonstationary signals, which may contain impulsive components and be nonuniformly sampled. The proposed methods significantly extend the conventional in two aspects: 1) nonstationarity problem is addressed by employing weighted least squares (WLS) to estimate locally time-varying an intersection confidence interval technique adaptively select window sizes WLS domain. yields adaptive WLP (AWLP) having better tradeoff between time resolution frequency resolution. 2) A more general regularized maximum-likelihood-type (M-) estimator used instead LS estimating AWLP. novel M-estimation-based AWLP method capable reducing estimation variance, accentuating predominant components, restraining adverse influence separating components. Simulation results were conducted illustrate advantages over resolution, sparse representation sinusoids, robustness applicability sampled data. Moreover, as computation at each sample independent others, parallel computing can conveniently employed without much difficulty reduce computational our real-time applications. expected find wide range applications instrumentation measurement related areas. Its potential power quality speech signal are also discussed demonstrated.