Paddy is one of the important long stored grains. It is usually stored in granaries to ensure continuous supply throughout the year, which can be subjected to deterioration due to various environmental conditions. Electronic nose (E-Nose) is used for non-destructive rapid in-situ detection of paddy for qualitative and quantitative assessment. Further, soft computing techniques are a promising solution for data analysis to enhance prediction accuracy. In this work, paddy quality assessment has been carried out using an E-Nose. Six different metal oxide semiconductor gas sensors are used for detecting simple and assorted volatile organic compounds evolving from paddy when stored over a period of time at different conditions. A Fuzzy Set based Multiple Linear Regression (FSMLR) has been proposed and implemented for paddy quality assessment. In the first stage, four fuzzified sensor data are given as input for assessment. While in the second stage, piecewise multiple linear equations were used for quality assessment and defuzzification. Quantitative assessment was carried out for nineteen multiple linear regression models built using fuzzy set theory. The proposed model achieves better performance metrics in terms of low root mean square error for cross validation and relative prediction error (RMSECV <0.028, RPE <1 × 10−4) for all the models. Further, adjusted regression coefficient (R2 = 0.99) confirms the good fit and Karl-Pearson coefficient of 0.99 confirms the prediction accuracy of the proposed model. This model can be embedded with sensor nodes for online monitoring of paddy granaries. © 2017 Elsevier B.V.