Abstract Tea taste assessments generally rely on panel sensory evaluation, which often yield inconsistent results. Therefore, the rapid and nondestructive assessment of the taste attributes of tea is important for its quality evaluation. This study assessed black tea taste attributes using a novel low-cost evaluation method that employed a smartphone-connected micro-near-infrared (micro-NIR) spectrometer. Bitterness and astringency intensity were evaluated by a trained panel, and caffeine and epigallocatechin gallate (EGCG) contents were analyzed using high-performance liquid chromatography. Partial least squares regression and multiple linear regression models were established on characteristic wavelengths selected using the successive projection algorithm and competitive adaptive reweighted sampling (CARS), respectively. The optimal prediction models obtained after conducting CARS selection yielded satisfactory results, with residual predictive deviation of 3.07, 2.28, 3.29, and 2.91 for bitterness score, astringency score, caffeine, and EGCG content, respectively. The results proved that micro-NIR spectrometers can be used to predict the taste attributes of black tea, providing a new method for the quality assessment black tea.

中文翻译：

---

用微近红外光谱技术评估红茶的口味相关属性

摘要 茶味评估通常依赖于小组感官评估，这往往会产生不一致的结果。因此，快速无损地评价茶叶的味觉属性对其品质评价具有重要意义。本研究使用一种新颖的低成本评估方法评估红茶的口味属性，该方法采用连接智能手机的微近红外 (micro-NIR) 光谱仪。苦味和涩味强度由经过培训的小组评估，并使用高效液相色谱分析咖啡因和表没食子儿茶素没食子酸酯 (EGCG) 的含量。分别使用逐次投影算法和竞争自适应重加权采样（CARS）选择的特征波长建立偏最小二乘回归和多元线性回归模型。进行CARS选择后得到的最优预测模型取得了令人满意的结果，苦味评分、涩味评分、咖啡因和EGCG含量的残余预测偏差分别为3.07、2.28、3.29和2.91。结果证明，微近红外光谱仪可用于预测红茶的口感属性，为红茶的品质评价提供了一种新的方法。