An Improved Classification of Pork Adulteration in Beef Based on Electronic Nose Using Modified Deep Extreme Learning with Principal Component Analysis as Feature Learning

Rapid meat consumption for human diets is causing a considerable production demand to fulfill. Due to many factors, such as economic problems, the adulteration problem of mixing meat with low-quality prices rises and affects the quality of food, bringing the issues to the health aspect. Therefore, it is necessary to ensure that the meat quality can be maintained by guaranteeing the process and the quality. The electronic nose (e-nose) is proposed to guarantee the process. However, recent research found that e-nose still suffers an issue of requiring a specific amount of time or several cycles to classify the meat sample. It happened because a feature extraction requires accumulating the data before extracting the signal into particular characteristics. It may affect the accumulation of processing time if repeated testing for meat samples. Therefore, this study proposes a new approach to accelerate classifying by removing the feature extraction block and applying the deep extreme learning machine (D-ELM). The powerful D-ELM method is then combined with principal component analysis (PCA) to classify pork adulteration in beef. The experiment shows that the proposed model achieved an accuracy of 99.97% in the tenfold cross-validation to classify seven mixed adulteration meat. In addition, the proposed method required within reach average execution time for model 4.014 s (one cycle after sensor sent signal) compared to previous research that needs more than 60 s (60 cycles) for evaluating a sample.