Rapid freshness prediction of crab based on a portable electronic nose system

In this paper, an automatic freshness prediction system for the living Chinese mitten crab was explored, which was formed from an electronic nose based on seven metal oxide semiconductor sensors. Prediction system acquired test data from the characteristic compounds in the headspace of crabs, and then was dealt with four different dimension reduction algorithms including PCA, LDA, KPCA and LE to reduce dimensions and extract effective features of sensor scores. Experimental results illustrated that the prediction system sensitively responded to crabs. PCA and LDA results failed to differentiate the response data of the living crabs. LE and KPCA were able to identify the different response data of crab samples. Back propagation neural network was used as a prediction model after dimension reduction. The model based on LE-BPNN reached a high identification rate of 90.6%. The simulation and experiment results showed that the prediction system can estimate the freshness of the living crab.