The paper presents an automatic recognition system for Amazigh characters using images from mobile phones, employing random forest and support vector machine (SVM) methods for classification. It outlines the methodology involving stages of pre-processing, feature extraction, and classification, with a dataset of 3,300 handwritten samples analyzed for performance. Experimental results demonstrate that the random forest classifier outperforms SVM with recognition rates of 97.75% and 93.17%, respectively, using the zoning feature extraction method.