Rapid assessment of meat quality by means of an electronic nose and support vector machines

Minced pork was stored aerobically and in MAP conditions at five different temperatures (0, 5, 10, 15, and 20 degrees C) and microbiological analysis in terms of total viable counts (TVC) was performed in parallel with e-nose measurements and sensory analysis until spoilage was evident in the samples. The volatile patterns collected from e-nose were initially subjected to Principal Component Analysis (PCA) for dimensionality reduction and subsequently to Support Vector Machines (SVM) analysis, using different kernels (linear, polynomial, and radial basis function), in order to classify meat in three distinct quality classes namely, fresh, semi-fresh, and spoiled. Results showed that SVM with radial basis function kernel provided good discrimination of minced pork samples regarding spoilage status. The overall correct classification in the three sensory classes was 81%, whereas correct classification for fresh, semi-fresh and spoiled samples amounted to 76, 87, and 78%, respectively.