Abstract: In order to predict freshness of different proportion of fat and lean pork,TVB-N content and nutrient of fresh pork was detected under the condition of 4℃.The volatile odor information of fresh pork was also detected by electronic nose technology.The regression prediction model of nutrient components was established with the characteristic values of sensor array.Two kinds of TVB-N neural network prediction models were established for classifying and not classifying the proportion of fat and lean.The results showed that the classification and establishment of neural network model to predict the effect better. After classifying the samples into two categories and establishing 2 models,correlation coefficient of the model training group was 0.994,0.985(p<0.01),the correlation coefficient of prediction group reached 0.984,0.979(p<0.01).The absolute error of the model was small and the distribution interval was concentrated. The 86% and 62.6% samples of the absolute error between 0~1 in the training group.There was no absolute error of more than 2.5 samples,only 8.5% samples in the prediction group was greater than 2.5.There exists a good correlation between e-nose sensor signal and TVB-N content,the electronic nose detection technique could be used as a rapid way to predict TVB-N content and to evaluate pork freshness with non-destructive test.