Plastic solid waste management is a growing field in many countries that is supported by governmental legislation. Plastic recycling is widely concerned since the other methods such as, landfilling and burning, harm the environment and are economically inefficient. Different methods have been proposed in literature based on the physical and thermal properties of the plastic but they are labor intensive and time consuming. The most promising method is based on the optical spectral sensing in the infrared range. It is non-destructive and time-efficient especially when coupled with machine learning models such as chemometrics. However, the infrared spectrometers are bulky, expensive and require building a model for each unit. Micro-electro-mechanical-system (MEMS) Fourier-transform infrared spectroscopy (FTIR) spectrometers are designed to be portable and lower in cost and in power consumption than traditional spectrometers. In this work, we present a classification predictor of two plastic categories Polyethylene Terephthalate (PETE) and Polypropylene (PP) using a portable MEMS FTIR spectrometer. PETE and PP samples are collected with different shapes and surface curvatures and the samples diffuse reflectance spectra are measured at 10 different surface spots per sample. The measurements are averaged and processed to mitigate the physical scatter effects. A multivariate linear classification chemometrics model is built and validated using one sample out cross validation. The classification using one principle component reaches a classification success rate of 100 % opening the door for low cost portable device for accurate plastic sorting.