Extracting proper features to understand the dynamics from the observed time series is a challenging problem, in particular in the modern era. Synchrosqueezing transform has been proved to be a mathematically rigorous and practically useful time frequency analysis technique for this purpose. The robustness property of synchrosqueezing transform has also been well studied in the past. However, it is well know that the "variance" induced by the inevitable noise could lead to several artificial speckles which misleading the interpretation. To alleviate this trouble, the idea of high dimensional geometric property of random projection is applied to design a new algorithm called ConceFT based on the synchrosqueezing transform. Its theoretical study will be discussed, and its application to anesthesia dataset will be shown.