Focusing light into an arbitrary pattern through multimode fiber is highly desired in energy delivery-related biomedical applications and has been demonstrated feasible with wavefront shaping. Here, the strategy relying on natural gradient ascent-based parameter optimization is shown to search the optimal wavefront rapidly towards high-quality pattern projection through a multimode fiber. Meanwhile, a new fitness function based on cosine similarity is adopted to replace the commonly used Pearson correlation coefficient, which leads to higher pattern contrast without sacrificing the fidelity with the target. With the proposed scheme, long-distance projection of arbitrary pattern was demonstrated through a 15-meter unfixed multimode fiber, showing fast convergence and better anti-interference ability. The superior performance of our approach for generating arbitrary pattern may gains special interest in multimode fiber based deep-tissue photon therapy and optogenetics.
|