Feature extraction coupled with electronic nose for Chinese mitten carb freshness identification

Identification of Chinese mitten crab freshness needs to establish an effective and special quality evaluation method. However, it is difficult to identify the quality for the complicated and variable physiological parameters of live crabs. In this paper, a method to rapidly determine living Chinese mitten crab freshness using an electronic nose (E-nose) and a Kernel-based technique is studied. Specifically, a novel kernel slow feature discriminant analysis (SFDA) as a nonlinear dimension reduction technique is adopted for feature extraction on the collected original data from E-nose to achieve the purpose of dimension reduction and denoising, In terms of support vector regression (SVR) learning technique, is used for modeling quality changes of crab during storage. Experiment shows that the regression model for crab TVB-N index, comparing with PCA and kernel PCA, kernel SFDA has higher accuracy with less training and testing. Therefore, the proposed approach can be used as an alternative tool for rapid detection of crab freshness.