Freshness evaluation of grouper fillets by inexpensive e-Nose and spectroscopy sensors

Abstract

Estimating the freshness of fish is a crucial aspect linked to its overall quality. In our research, a rapid freshness sensing system was designed for predicting the freshness of grouper (Epinephelus coioides) fillets throughout their storage period. This system incorporated ten gas sensors (assembled as e-nose) and 18-wavebands spectroscopy chip. Grouper fillet each was placed inside of the sensing device and stored at a temperature of 10 +/- 2degree celsius for a duration of 72 h with sensor signal acquired hourly. In total 292 data points were collected. The true freshness of the grouper fillets represented by K value and total volatile basic nitrogen (T-VBN), was determined by HPLC and colorimetric method. Machine learning algorithms, including multiple linear regression (MLR), partial least squares regression (PLSR), and support vector regression (SVR), were employed to construct predictive models based on the data form both sensors and instrumental analysis. Among these models, MLR observed superior performance in predicting K value and T-VBN, achieving root mean square errors (RMSE) of 5.959 and 1.473, along with R2 of 0.914 and 0.932, respectively. The sensing system exhibited great predicting capabilities for assessing fish freshness.