A handwritten signature is one of the behavioral biometric traits of an individual, which is commonly used for authentication and document validation in various organizations. The signature patterns of an individual vary with age, time, mood, emotion, illness and other environmental factors, which makes signature recognition a challenging task. Another challenge is low inter-class variability of signatures since the forgers imitate the signatures of others for unauthorized access. To overcome those challenges, many researches were conducted in signature recognition using machine learning and deep learning approaches. However, there are still gaps that need to be investigated, such as the overfitting problem, computational complexity, poor preprocessing, and scalability problem. Hence, the aim of this research is to develop an offline handwritten signature recognition model with better performance in dealing with the specified gaps. We have developed three different models (custom CNN, CNN+SVM, HOG+SVM) and ensemble them using the weighted average voting ensemble learning technique. In order to carry out the experiment, we used a custom signature image dataset that was collected from 60 different individuals. From each individual, we have collected 40 samples with different orientation angles to corporate more intra-class variability of the signature. The signature images were preprocessed using techniques like grayscaling, resizing, normalization, augmentation, thinning, and filling disconnected patterns. After preprocessing we used a custom CNN model and HOG descriptor as feature extractor to develop CNN+SVM and HOG+SVM models. After training and evaluation of each model, the three models were ensembled together using weighted average voting ensemble technique. Therefore, we evaluated the ensembled model, and we got an accuracy of 99.58% which outperforms the individual model. Our system outperforms the related work performance while compared to the existing research work. Keywords: ML, CNN, Signature Recognition, Ensemble learning