Cardamom Quality Evaluation Employing Electronic Nose

EasyChair Preprint 7272

Abstract

Cardamom is an export-oriented commodity of India, used for cooking and has medicinal values. This calls for the need of rapid and effective method for quality assessment of cardamom. The major parameters for quality determination of cardamom are flavour and aroma. The present practice of quality estimation involves assessment by human experts based on physical characteristics, such as freshness, shape, size, colour and aroma which is subject to biasness and error. The alternate evaluation technique comprises of various qualitative and quantitative chemical tests of the samples (using gas chromatography and mass spectrometer), which is an expensive, labourious, time-consuming and skilled man-power demanding process. This paper presents a novel approach to predict the oil yield and constituent chemicals 1,8-cineole and alpha-terpinyl acetate, which are responsible for the flavour of cardamom using a handheld electronic-nose (HEN) developed by Centre for Development of Advanced Computing (CDAC), Kolkata. We worked with thirteen different cardamom samples with varying oil yield percentages and different measures of the constituent chemicals of the extracted essential oil. We considered cardamom samples, with and without husk separately in our study. Initially, we have applied clustering algorithms like Principal Component Analysis (PCA) and Density-Based Spatial Clustering on Applications with Noise (DB-SCAN) on the data values collected by five different metal oxide semiconductor (MOS) sensors of HEN and the device was able to group the samples into distinct clusters with considerable accuracy. Then, Partial Least Square (PLS) Regression was applied on the dataset to train the system and eventually predict the quality of unknown cardamom samples. The model has given around 98% and 95% accuracy for oil yield and 1,8-cineole prediction respectively for the samples with husk.

Keyphrases: Cardamom, Density-Based Spatial Clustering of Applications with Noise, Metal Oxide Semiconductor Sensor, Partial Least Square Regression, Principal Component Analysis, electronic nose, quality evaluation

@booklet{EasyChair:7272,
  author    = {Madhurima Ghosh and Devdulal Ghosh and Nabarun Bhattacharyya},
  title     = {Cardamom Quality Evaluation Employing Electronic Nose},
  howpublished = {EasyChair Preprint 7272},
  year      = {EasyChair, 2021}}