A data-driven meat freshness monitoring and evaluation method using rapid centroid estimation and hidden Markov models

Publication Type:
Journal Article
Sensors and Actuators, B: Chemical, 2020, 311
Issue Date:
Filename Description Size
1-s2.0-S092540052030215X-main.pdfPublished version4.63 MB
Adobe PDF
Full metadata record
© 2020 Elsevier B.V. The food industry needs a fast and stable method to monitor and evaluate the freshness of meat products. This paper proposes the use of a metal-oxide-semiconductor (MOS) sensor-based electronic nose with a data-driven approach, hidden Markov models (HMMs), to detect meat spoilage. The data-driven approach we proposed include: (i) A training algorithm using segmental rapid centroid estimation (RCE) to increase the overall stability of the HMM optimisation algorithm; (ii) A monitoring algorithm using a single HMM trained only by fresh samples to track the change of meat freshness in real-time when the storage condition is not specified; (iii) An HMM-based decoding algorithm to cluster the freshness level when the whole life-span data of meat stored in a specific storage condition is available. Then, HMM for each freshness level is trained and applied in parallel as freshness evaluation models to classify the tested meat samples. In freshness monitoring, case studies on fish, beef and chicken demonstrate the effectiveness and potential of this method. In freshness evaluation, we observed sensitivity and specificity of 96.32% ± 2.83% and 99.07% ± 0.69%, 99.09% ± 2.40% and 99.82% ± 0.48%, as well as, 99.35% ± 0.27% and 98.31% ± 0.71% respectively for fish, beef and chicken, which warrants further investigation.
Please use this identifier to cite or link to this item: