Deep learning in a sensor array system based on the distribution of volatile compounds from meat cuts using GC-MS analysis

Generally, people distinguish the type of meat by looking at the color, texture, and even aroma of meat. These
three methods have less effective approaches to distinguish the types of meat from meat cuts. Some researchers
analyze the differences in the aroma of meats by using laboratory equipment, which is gas chromato-
graphy–mass spectrometry (GC–MS). This tool is mostly accurate, but it requires some time to determine the
meat types completely. Moreover, the analysis process using GC–MS is also complicated. Nowadays, the elec-
tronic nose (e-nose) is a promising technology because it has a faster process of identifying various food types
with reasonable production costs. Hence, the development of an e-nose for distinguishing volatile compounds
from some meat types is appealing. Not only to determine the type of meat, but this study can also differentiate
the part of the body from the meat, which has never been done by previous researchers. GC–MS was used as
ground truth for the e-nose system, which helped the results to meet the standards. To achieve the objective in
differentiating two meat cuts from three types of meat, this study uses statistical parameters for extraction
feature, PCA for reducing the dimension, and deep learning.
Furthermore, to get more improvements from the previous researches, this study aims to optimize the
parameters of deep learning. The result of the proposed method was compared to several machine learning
algorithms that were used in previous studies, i.e., k-nearest neighbor (k−NN), support vector machine (SVM),
Multi-Layer Perceptron (MLP), and basic deep learning. The experimental results showed that e-nose could
detect meat cuts for 120 s, and the proposed method provides a significant improvement.