Feature-based image classification is a commonly used approach in human identification for multimodal biometric systems. It involves extracting distinctive features from images, such as facial features, finger veins, and iris texture. These features are used to classify the images into different classes. Additionally, researchers are exploring the use of multi-modal biometric systems, which combine multiple sources of information, such as face and iris recognition for improved accuracy as compared to the other multi-traits. The current state-of-the-art approaches in feature-based image classification for human identification in multimodal biometric systems aim to increase accuracy, reduce error, and enhance user convenience. In this survey, the principles and restrictions of unimodal identity recognition are analyzed. Apart from this, the motivations and benefits of multimodal identity recognition are discussed. The performance of various techniques such as Convolutional Neural Network (CNN), Support Vector Machine(SVM), Hybrid Localized model, and deep learning algorithms used by different researchers have been discussed analytically on the different modalities. Hybrid Localized Convolutional Neural Network (HLCNN). Compared to the various techniques HLCNN and deep learning have provided the best accuracy. Moreover, it is found that the accuracy of the multi-modal biometric can be enhanced by using more than two or three biometric traits. In this research, the datasets are also explained which tells us about what type and number on which the research has been conducted.