Current advances in technology have highlighted the importance of video analysis in the domain of computer vision. Traditional artificial neural networks have considerably high computational costs with video analysis, and many modern applications such as autonomous vehicles have limited computational resources. Spiking neural networks (SNNs) are third generation, biologically plausible models that are seen as hypothetical solutions for the bottlenecks of ANNs, such as energy efficiency. However, current SNN-specific methods that achieve good classification rates, such as ANN-to-SNN conversion and back-propagation, depend on labeled data, which requires costly human intervention. Meanwhile, unsupervised learning with SNNs using the spike timing-dependent plasticity (STDP) rule has the potential to overcome some bottlenecks of regular artificial neural networks. However, STDP-based SNNs are still immature. SNNs trained in an unsupervised manner with STDP can hypothetically surpass ANNs in energy efficiency, and thus must be studied and improved. In this work, we study the performance of these networks with human action recognition tasks. Moreover, we focus on the motion found in videos in order to recognise the actions. In this paper, we focus on studying the effects that different motion modeling techniques can have on the spatio-temporal features extracted by a spiking neural network trained with unsupervised STDP.