Ultrashort pulse identification based on a temporal-frequency dual-domain recognition algorithm

The automatic identification of pulse types generated by ultrashort pulse lasers represents a critical advancement in laser science and precision measurement, facilitating the transition from "experience-driven" to "data-driven" intelligent laser systems. This study systematically investigates the dynamic characteristics of soliton pulses in passive mode-locked fiber lasers through comprehensive numerical simulations, focusing on the effects of intracavity net dispersion and saturable absorption energy on conventional solitons, self-similar pulses, dissipative solitons, and multiple solitons. To enhance the identification efficiency of different soliton types, we propose, to our knowledge, a novel temporal-frequency dual-domain recognition algorithm. The algorithm employs a two-stage approach: the first stage accurately determines the number of solitons based on temporal domain features, while the second stage classifies individual soliton types through spectral feature analysis. When applied to soliton identification across a two-dimensional parameter space, the algorithm successfully identifies 1435 sets of output solitons within 9 s, demonstrating exceptional speed and accuracy. This work not only advances our understanding of soliton distribution patterns in parameter space but also establishes a foundation for intelligent laser control and parameter optimization.