A 2015 study from Japan looking at two groups of English learners found that indeed, Shadowing made a statistically significant improvement in phoneme perception for both groups.
But, even from Day 1, it can be a super efficient way for improving your pronunciation and at the same time your ability to recognize phonemes in natural language.
Statistical speech models and a probabilistic technique called Gaussian mixture modelling are then used to identify each phoneme, before reconstructing the original word.