Understanding time series is critical for a wide variety of fields in today’s data driven world. For tasks involving irregular data with variability that can obscure larger scale trends, such as classifying human interaction, the current state-of-the art can fall short. Topological data analysis (TDA) is effective at identifying structure in the face of noise, but it is not clear whether its demonstrated strengths on higher dimensional data transfer over to unidimensional temporal data. In this paper we first demonstrate that TDA is a capable technique for time series classification, and subsequently that it performs notably well on noisy human generated data.