Tactile object recognition is crucial for robots operating in environments where visual information is unreliable. While traditional machine learning approaches for tactile object recognition often struggle with noise and sensor variations, persistent homology, a tool from topological data analysis, offers a robust representation of object shape across different scales. This paper explores the application of ideas from Topological Data Analysis, specifically, persistent homology to enhance the noise resilience of tactile object recognition. We demonstrate how persistent homology features, specifically persistent entropy, can be extracted from tactile images and combined with traditional features for object classification. Through experiments on a tactile image dataset \cite{cnn_gandarias}, we present exploratory results on the performance of several sklearn classifiers with and without persistent entropy as a feature, showcasing the improved robustness achieved through the inclusion of topological information.