Age detection of mature vinegar based on electronic tongue and electronic nose combined with DenseNet-ELM

Abstract

Objective: In order to realize the rapid detection of the brewing age of mature vinegar. Methods: Electronic tongue （ET） and electronic nose （EN） combined with Densely Connected Convolutional Networks-Extreme Learning Machine （DenseNet-ELM） model were used to quickly detect the brewing age of mature vinegar. Two DenseNet models with different structures, ET-DenseNet and EN-DenseNet, were designed to extract the feature information of the electronic tongue and electronic nose signals respectively. And then the feature level information fusion method was used to obtain the fusion feature vectors of the two artificial sensory devices. Then Extreme Learning Machine （ELM） was used to classify and recognize the fused feature vectors. Results: DenseNet can effectively extract the deep features of electronic tongue and electronic nose signals, and its feature extraction ability was better than Discrete Wavelet Transform （DWT） and Convolutional Neural Network （CNN）; Compared with the use of electronic tongue or electronic nose alone, the information fusion method had better accuracy and robustness for the detection of mature vinegar of different years. The Accuracy, Precision, Recall and F1-score of the test set reach 99.1%, 0.98, 0.99 and 0.99, respectively. Conclusion: The dense convolution network can alleviate the problems of model degradation and weak generalization ability caused by the increase of depth of the deep learning model, and can effectively classify seven kinds of aged vinegar with different brewing years.

Keywords

• mature vinegar
• age detection
• electronic tongue
• electronic nose
• information fusion