<p>Limited by the stability of characterization indexes and the evaluation capability for repeated transient pulses, the performance of conventional blind deconvolution methods in extracting repeated transient components has not seen substantial improvement. To address this limitation, this study builds upon previous research by defining a novel index termed the generalized mixed-domain harmonic-to-noise ratio (GMHNR). Based on this index, we propose an enhanced blind deconvolution method specifically designed for accurate identification of repeated transient pulses, named the blind deconvolution method induced by GMHNR (BD-GMHNR). Notably, the GMHNR index incorporates an expansion factor that effectively mitigates the impact of non-strict periodicity inherent in actual fault characteristics. Furthermore, the index employs generalized Ramanujan spectrum (GRS) analysis to optimize frequency domain distribution. This approach not only endows the index with robust cyclostationary properties but also ensures more stable evaluation capability through strict monotonicity. The strict monotonic nature of the GMHNR index guarantees that the BD-GMHNR method can precisely identify optimal filtering results, preventing convergence toward suboptimal solutions and thereby ensuring accurate extraction of repeated transient pulses. Both simulation studies and experimental signal analyses demonstrate that the proposed BD-GMHNR method exhibits superior performance in extracting repeated transient components, making it particularly suitable for engineering applications such as bearing fault diagnosis.</p>