Abstract: Abstract: The ripeness stage is one of the important factors of banana quality, which will affect the sale price of bananas. To consumers, inadequate ripeness means low quality, and over-ripeness means a short shelf life, and both of them will lead to low income to sellers. In order to obtain optimal quality, keep a long shelf period, and minimize losses, it is necessary to automatically detect banana ripeness nondestructively. One promising strategy is using a quartz crystal microbalance (QCM) system to detect volatile organic compounds (VOCs) released during banana ripening. The advantages of this method are quick detection, high sensitivity, and low cost.A QCM sensor is a mass sensor, which is a device of an oscillating circuit. When the mass on its surface is changed, the frequency of an oscillating circuit will change simultaneously. When coated by cross-sensitivity materials, which can absorb VOCs. QCM sensors can be used to detect the change of ambient VOCs. Therefore, it is possible to use QCM sensors for recognizing the banana ripeness stage.This research aimed to discriminate the ripeness stage of banana based on QCM sensors. The measurement system was constructed with 9 MHz of gold electrodes quartz crystals, which were modified by sensitive materials of 1, 2-dioleoyl-sn-glycero-3-phosphor-L-serine (DOPS), galactosylceramide (GC), cellulose acetate (CA), and ethyl cellulose (EC). The sensitive materials were dissolved in chloroform (CHL) and tetrahydrofuran (THF). The QCM sensors were fabricated by dropping 5 mg/mL DOPS/CHL, GC/CHL, and CA/THF, EC/THF on quartz crystal electrodes, respectively. Bananas were purchased from the Philippines in green condition, and they were ripened in a room with a constant temperature of (22±1)℃. During ripening, the bananas were divided into 7 ripening stages according to a Von Loesecke H W chart, the banana VOCs were obtained by static head-space through putting bananas into a 2000 mL teflon chamber for 1 hour at every ripening stage. The measurement was conducted by injecting 50 mL VOCs into a 150 mL measurement chamber. The response of these 4 sensors to VOCs of bananas from the 1 to 7 ripening stage was recorded by LabVIEW software. The experiments were repeated 8 times, and the response value was analyzed by principal-components analysis (PCA) and linear discriminant analysis (LDA).The results showed that the response of a sensor modified by CA was more sensitive to banana VOCs than DOPS, GC, and EC. For the sensor coated by CA, the response tended to increase with the ripening stage from stage 1 to stage 4, the frequency shift reached its peak at ripening stage 4 due to an increase of aldehydes and esters in the banana VOCs. After ripening stage 4, the frequency shift declined with a decrease of acetates and an increase of butyrates and alcohols. The other three sensitive materials modified sensors' response were lower than CA, though their trends were not the same. The frequency shift of 4 sensors coated by 4 sensitive materials to banana VOCs of 7 ripening stages was classified by PCA and LDA, with accuracies of 97% and 100%, respectively. The research showed that it was feasible to classify the ripening stage of banana using four materials as sensitive film, and that the LDA is a potential classification method. Accordingly, this research is helpful for banana automatic grading.