The pork freshness is a big safety issue on people's health.In this paper,fresh pork samples were stored in decompression storage room.The TVB-N content,total bacterial count,pH value and sensory scores of the samples were determined at different storage stage.SVM neural networks models were obtained by training the sample data with different kernel functions and cross-validation.Furthermore,the test data were used to predict the freshness of pork sample by SVM neural network.The experiment results suggested that the SVM neural networks obtained higher correct classification rate of pork freshness with the right kernel function and cross-validation according to the sample performance.