In the 2024 Indonesian Presidential Election, the digit recognition system for vote recapitulation faced serious challenges due to data input errors at several polling stations (TPS), affecting the accuracy of the results. The research data was obtained from the fourth series dataset containing images of handwritten digits, extracted from a zip file and processed into 64x64 pixel images. The research method involved feature extraction using Histogram of Oriented Gradients (HOG) to detect edges and gradient directions in the images, and pattern recognition using the K-Nearest Neighbors (KNN) and Support Vector Machine (SVM) algorithms. Experimental results showed that the combination of HOG and KNN achieved accuracies of 91.35%, 91.65%, and 92.25% for training and testing data ratios of 60:40, 70:30, and 80:20, respectively, while the use of HOG with SVM achieved accuracies of 97.48%, 97.48%, and 97.68% for the same ratios. Without feature extraction, KNN accuracy reached only 40.79%, 41.76%, and 42.35%, while SVM reached 89.99%, 89.75%, and 89.79% for the same ratios. From these experimental results, it can be concluded that the combination of HOG feature extraction and the SVM algorithm provides the best results, with accuracies exceeding 97% across various dataset proportions. This underscores the importance of applying appropriate feature extraction techniques to enhance the performance of pattern recognition systems, particularly in the context of handwritten digit recognition for the 2024 Presidential Election.