Research On Chicken Flavor Recognition And Correlation Model

Chicken essence is one kind of composite seasoning with rapid growth of fresh ingredient, and the flavor quality assessment relied on artificial sensory evaluation or analysis ofelectronic equipment.Artificial sensory evaluation is a scientific way to measure, analysis and explain the product by human vision, smell, taste, etc. Generally the food color, flavor type, texture type and others are selected as artificial taste sensory indicators. But the human senses are affected by evaluators physical and psychological factors, also qualitative changes of substance cannot suit for artificial tasting, so the electronic equipment to aid in sensory evaluationElectronic equipment, including electronic nose and electronic tongue, is simulated human sensory response function to generate the fingerprint pattern of the test sample, and used multivariate statistical methods to obtain classification results, which have fast and accurate advantages. However, multivariate statistical analysis is the study of the statistical laws of objective things interdependence among multiple variables, if the sample type and number are huge, the statistics result is inaccurate, even a big error. In addition, electronic equipment collected the data which is taste information integrity to distinguish samples, and identified the degree of similarity of samples, but cann't obtain each evaluation differences.The correlation research between artificial sensory and electronic sensory, focused on the correlation analysis between sensory index and sensor variables, but the correlation research of two chicken sensory evaluation system is very little.This paper presents the following two solutions for the inadequate above evaluation of chicken flavor seasoning:First, the object of study are chicken samples in six different ingredients added, and collected electronic noses'fingerprint data, then established classification model by principal component analysis, clustering, standard BP neural network algorithm, the improved BP neural network algorithm and principal component analyse combined with improved BP neural network algorithm, then compared the classification results to found, the chicken flavor quality classification prediction model created by the principal component analysis and improved BP neural network algorithm, which not only classify chicken samples qucikly and accurately, also detected the flavor does not meet predetermined requirements samples.Second, the research object are chicken samples of three different shelf life, and acquired artificial taste evaluation data and electronic tongue fingerprint data, then established the correlation model between evaluation of artificial taste and electronic tongue fingerprint data by the improved BP neural network, which can classified unknown samples, and given the unknown sample's artificial sensory data directly by electronic tongue data, then compared the differences of artificial sensory evaluation index between standard sample and unkown sample in this category. The model not only meet the objective and fast reqiurement in chicken sensory evaluation process, but also given the differences information in chicken quality verification process.