Crispness is one of the most important indexes for Ctenopharyngodon idella. Over-crispy C. idella will show hemolysis, hypoxia and pathological changes of some organs. Therefore, it is urgent to develop method to detect the crispness of C. idella. In this paper, a method based on Raman spectroscopy was proposed to detect the crispness of C. idella. Firstly, the Raman spectra of C. idella with different crisping time were analyzed by PCA method. The results showed that Raman spectroscopy could be used to identify the crispness of crispy C. idella with different crisping time. Additionally, with the extension of the crisping time, the α-helix of the total protein in the muscle of crispy C. idella was decreased, while the β-fold of the total protein in the muscle of crispy C. idella was increased. In terms of irregular curl of the total protein in the muscle of C. idella, obvious changes were observed at the early stage, but not at the late stage of crisping. Secondly, four preprocessing methods, including SG, SNV, MSC and Normalize, were used to preprocess the Raman spectrum data. It was found that Normalize method had the optimal preprocessing effect, with RMSEP of 2.33 and R²_P of 0.73. Thirdly, PLSR, SVR and BPNN were used to establish the relationship model between crispness and Raman spectrum information from the muscle of C. idella. The RMSEP of the prediction set was 2.33, 2.26 and 1.96 , respectively, and the R²_P of the prediction set was 0.73, 0.78 and 0.83, respectively. Apparently, the BPNN model was the optimal one. Taken together, our results showed that the Normalized-BPNN prediction model based on Raman spectroscopy could effectively detect the crispness of C. idella. This study paves a new way for C. idella muscle crispness detection methods in the future.