Most of the work done in the field of Optical Music Recognition (OMR) is based on perfectly digitalized music scores or hand written ones as input for various algorithms developed with the goal to translate them into a language a machine can understand. However, when it comes to a mobile environment, external factors such as exposure to the elements play a huge role in the acquisition of the images. The preprocessing stage requires more attention in order to prepare the images to be classified and the classification stage has to take as little time as possible without affecting the results since we aren't working with desktop grade processing speeds. This work presents a comparative analysis between Support Vector Machine (SVM), Sequential Minimal Optimization for SVM (SMO), Multilayer Perceptron (MLP), Random Trees and Naive Bayes algorithms in the classification of whole notes, half notes, quarter notes and eight notes. This analysis is focused on the accuracy and time required to train the dataset for each classifier.